Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
~91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
METHOD AND APPARATUS FOR CURRENCY
DI8~PTMTN~TION AND COu.~lNG
BA~RGR~UND OF THE l~.v~:~.,lON
~ield of the Invention
The present invention relates, in general, to
currency identification. The invention relates more
particularly to a method and apparatus for automatic
discrimination and counting of currency bills of
different denominations using light reflectivity
characteristics of indicia printed upon the currency
bills.
Description of the Related Art
A variety of techniques and apparatus have been used
to satisfy the requirements of automated currency
handling systems. At the lower end of sophistication in
this area of technology are systems capable of handling
only a specific type of currency, such as a specific
dollar denomination, while rejecting all other currency
types. At the upper end are complex systems which are
capable of identifying and discriminating between and
automatically counting multiple currency denominations.
Currency discrimination systems typically employ
either magnetic sensing or optical sensing for
discriminating between different currency denominations.
Magnetic sensing is based on detecting the presence or
absence of magnetic ink in portions of the printed
indicia on the currency by using magnetic sensors,
usually ~errite core-based sensors, and using the
detected magnetic signals, after undergoing analog or
digital processing, as the basis for currency
discrimination. The more commonly used optical sensing
technique, on the other hand, is based on detecting and
~" 91/11778 2 0 5 0 5 8 9 PCT/US91/00~'
~, ,
analyzing variations in light reflectance or
transmissivity characteristics occurring when a currency
bill is illuminated and scanned by a strip of focused
light. The subsequent currency discrimination is based
on the comparison of sensed optical characteristics with
prestored parameters for different currency
denominations, while accounting for adequate tolerances
reflecting differences among individual bills of a given
denomination.
A major obstacle in implementing automated currency
discrimination systems is obtaining an optimum compromise
between the criteria used to adequately define the
characteristic pattern for a particular currency
denomination, the time required to analyze test data and
compare it to predefined parameters in order to identify
the currency bill under scrutiny, and the rate at which
successive currency bills may be mechanically fed through
and scanned. Even with the use of microprocessors for
processing the test data resulting from the scanning of a
bill, a finite amount of time is required for acquiring
samples and for the process of comparing the test data to
stored parameters to identify the denomination of the
bill. Most of the optical scanning systems available
today utilize complex algorithms for obt~;n;ng a large
number of reflectance data samples as a currency bill is
scanned by an optical scanhead and for subsequently
comparing the data to corresponding stored parameters to
identify the bill denomination. Conventional systems
require a relatively large number of optical samples per
bill scan in order to sufficiently discriminate between
currency denominations, particularly those denominations
for which the reflectance patterns are not markedly
distinguishable. The use of the large number of data
samples slows down the rate at which incoming bills may
be scanned and, more importantly, requires a
correspondingly longer period of time to process the data
in accordance with the discrimination algorithm. The
WO91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
_ 3
processing time is further increased if the system is
also adapted to counting of identified currency
denominations. The speed at which bills can be
classified and counted is thereby restricted since real
time processing requires that the analysis of scanned
data for a bill be completed and the bill be identified
and counted as belonging to a particular currency
denomination before the subsequent bill gets positioned
across and is scanned by the scanhead.
A major problem associated with conventional systems
is that, in order to obtain the required large number of
reflectance samples required for accurate currency
discrimination, such systems are restricted to scanning
bills along the longer dimension of currency bills.
Lengthwise scanning, in turn, has several inherent
drawbacks including the need for an extended transport
path for relaying the bill lengthwise across the scanhead
and the added mechanical complexity involved in
accommodating the extended path as well as the associated
means for ensuring uniform, non-overlapping registration
of bills with the sensing surface of the scanhead.
The end result is that systems capable of accurate
currency discrimination are costly, mech~nically bulky
and complex, and generally incapable of both currency
discrimination and counting at high speeds with a high
degree of accuracy.
~UMM~RY OF THE l N v t;~. . lON
It is a principal object of the present invention to
provide an improved method and apparatus for identifying
and counting currency bills comprising a plurality of
currency denominations.
It is another object of this invention to provide an
improved method and apparatus of the above kind which is
capable of efficiently discriminating between and
counting bills of several currency denominations at a
high speed and with a high degree of accuracy.
`"~91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
A related object of the present invention is to
provide such an improved currency discrimination and
counting apparatus which is compact, economical, and has
uncomplicated construction and operation.
Briefly, in accordance with the present invention,
the objectives enumerated above are achieved by means of
an improved optical sensing and correlation technique
adopted to both counting and denomination discrimination
of currency bills. The technique is based on the optical
sensing of bill reflectance characteristics obtained by
illuminating a~d scanning a bill along its narrow
dimension, approximately about the central section of the
bill. Light reflected from the bill as it is optically
scanned is detected and used as an analog representation
of the variation in the "black" and "white" content of
the printed pattern or indicia on the bill surface.
A series of such detected reflectance signals are
obtained by sampling and digitally processing, under
microprocessor control, the reflected light at a
plurality of predefined sample points as the bill is
moved across the illuminated strip with its narrow
dimension parallel to the direction of transport of the
bill. Accordingly, a fixed number of reflectance samples
is obtained across the narrow dimension of the note. The
data samples obtained for a bill scan are subjected to
digital processing, including a normalizing process to
deaccentuate variations due to "contrast" fluctuations in
the printed pattern or indicia existing on the surface of
the bill being scanned. The normalized reflectance data
represent a characteristic pattern that is fairly unique
for a given bill denomination and incorporates sufficient
distinguishing features between characteristic patterns
for different currency denominations so as to accurately
differentiate therebetween.
By using the above approach, a series of master
characteristic patterns are generated and stored using
"original" or "new" bills for each denomination of
WO91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
currency that is to be detected. According to a
preferred embodiment, four characteristic patterns are
generated and stored within system memory for each
detectable currency denomination. The stored patterns
correspond, respectively, to optical scans performed on
the "top" and "bottom" surfaces of a bill along "forward"
and "reverse" directions relative to the pattern printed
on the bill. Preferably, the currency discrimination and
counting method and apparatus of this invention is
adapted to identify seven (7) different denominations of
U.S. currency, i.e., Sl, $2, ~5, Slo, S20, S50 and $100.
Accordingly, a master set of 28 different characteristic
patterns is stored within the system memory for
subsequent correlation purposes.
According to the correlation technique of this
invention, the pattern generated by scanning a bill under
test and processing the sampled data is compared with
each of the 28 prestored characteristic patterns to
generate, for each comparison, a correlation number
representing the extent of similarity between
corresponding ones of the plurality of data samples for
the compared patterns. Denomination identification is
based on designating the scanned bill as belonging to the
denomination corresponding to the stored characteristic
pattern for which the correlation number resulting from
pattern comparison is determined to be the highest. The
possibility of a scanned bill having its denomination
mischaracterized following the comparison of
characteristic patterns, is significantly reduced by
defining a bi-level threshold of correlation that must be
satisfied for a "positive" call to be made.
In essence, the present invention provides an
improved optical sensing and correlation technique for
positively identifying any of a plurality of different
bill denominations regardless of whether the bill is
scanned on its "top" or "bottom" face along either the
"forward" or "reverse" directions. The invention is
' ~ 91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
particularly adapted to be implemented with a system
programmed to track each identified currency denomination
so as to conveniently present the aggregate total of -
bills that have been identified at the end of a scan run.
Also in accordance with this invention, currency
detecting and counting apparatus is disclosed which is
particularly adapted for use with the novel sensing and
correlation ~e~hnique summarized above. The apparatus
incorporates an abbreviated curved transport path for
accepting currency bills that are to be counted and
transporting the bills about their narrow dimension
across a scanhead located downstream of the curved path
and onto a conventional stacking station where sensed and
counted bills are collected. The scanhead operates in
conjunction with an optical encoder which is adapted to
initiate the capture of a predefined number of
reflectance data samples when a bill (and, thus, the
indicia or pattern printed thereupon) moves across a
coherent strip of light focused downwardly of the
scanhead.
The scanhead uses an array of photo diodes to focus
a coherent light strip of predefined dimensions and
having a normalized distribution of light intensity
across the illuminated area. The photo diodes are
angularly disposed and focus the desired strip of light
onto the narrow dimension of a bill positioned flat
across the scanning surface of the scanhead. A photo
detector positioned above the illuminated strip detects
light reflected upwardly from the bill. The photo
detector is controlled by the optical encoder to obtain
the desired reflectance samples.
Initiation of sampling is based upon the detection
of the change in reflectance value that occurs when the
outer border of the printed pattern on a bill is
encountered relative to the reflectance value obtained at
the edge of the bill where no printed pattern exists.
According to a feature of this invention, illuminated
W091/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
-
strips of at least two different dimensions are used for
the scanning process. A narrow strip is used initially
to detect the starting point of the printed pattern on a
bill and is adapted to distinguish the thin borderline
that typically marks the starting point of and encloses
the printed pattern on a bill. For the rest of the
narrow dimension scanning following detection of the
border line of the printed pattern, a substantially wider
strip of light is used to collect the predefined number
of samples for a bill scan. The generation and storage
of characteristic patterns using "original" notes and the
subsequent comparison and correlation procedure for
classifying the scanned bill as belonging to one of
several predefined currency denominations is based on the
above-described sensing and correlation technique.
Brief Description Of The Drawings
Other objects and advantages of the invention will
become apparent upon reading the following detailed
description in conjunction with the drawings in which:
FIG. l is a functional block diagram illustrating
the conceptual basis for the optical sensing and
correlation method and apparatus, according to the system
of this invention;
FIG. 2 is a block diagram illustrating a preferred
circuit arrangement for processing and correlating
reflectance data according to the optical sensing and
counting technique of this invention;
FIGS. 3-8 are flow charts illustrating the sequence
of operations involved in implementing the optical
sensing and correlation technique;
FIGS. 9A-C are graphical illustrations of
representative characteristic patterns generated by
narrow dimension optical scanning of a currency bill;
FIGS. l0A-E are graphical illustrations of the
effect produced on correlation pattern by using the
--~91/11778 2 0 5 0 5 8 9 PCT/US9l/~ ~3
progressive shifting technique, according to an
embodiment of this invention;
FIG. 11 is a perspective view showing currency
discrimination and counting apparatus particularly
adapted to and embodying the optical sensing and
correlation tech~i que of this invention;
FIG. 12 is a partial perspective view illustrating
the merh~nism used for separating currency bills and
injecting them in a sequential fashion into the transport
path;
FIG. 13 is a side view of the apparatus of FIG. 11
illustrating the separation mechanism and the transport
path;
FIG. 14 is a side view of the apparatus of FIG. 11
illustrating details of the drive mechanism;
FIG. 15 is a top view of the currency discriminating
and counting apparatus shown in FIGS. 11-14;
FIG. 16 is a sectional side view showing the angular
disposition of the photo diodes within the scanhead;
FIG. 17 is an illustration of the light distribution
produced about the optical scanhead; and
FIG. 18 is a top view illustration of the optical
mas~ used to generate the scan strips of different
dimensions.
While the invention is susceptible to various
modifications and alternative forms, specific embodiments
thereof have been shown by way of example in the drawings
and will herein be described in detail. It should be
understood, however, that it is not intended to limit the
invention to the particular forms disclosed, but on the
contrary, the intention is to cover all modifications,
equivalents, and alternatives falling within the spirit
and scope of the invention as defined by the appended
claims.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
WO91/11778 PCT/US91/~ ~3
2050589
g
Referring now to FIG. 1, there is shown a functional
block diagram illustrating the optical sensing and
correlation system according to this invention. The
system 10 includes a bill accepting station 12 where
stacks of currency bills that need to be identified and
counted are positioned. Accepted bills are acted upon by
a bill separating station 14 which functions to pick out
or separate one bill at a time for being sequentially
relayed by a bill transport me~h~ni~m 16, according to a
precisely predetermined transport path, across an optical
cc~nhead 18 where the currency denomination of the bill
is ~cAnned~ identified and counted. The scanned bill is
then transported to a bill stacking station 20 where
bills so processed are stacked for subsequent removal.
The optical scanhead 18 comprises at least one light
source 22 directing a beam of coherent light downwardly
onto the bill transport path so as to illuminate a
substantially rectangular light strip 24 upon a currency
bill ~7 positioned on the transport path below the
scanh_ad 18. Light reflected off the illuminated strip
24 is sensed by a photodetector 26 positioned directly
above the strip. The analog output of photodetector 26
is converted into a digital signal by means of an analog-
to-digital (ADC) convertor unit 28 whose output is fed as
a digital input to a central processing unit (CPU) 30.
According to a feature of this invention, the bill
transport path is defined in such a way that the
transport mech~nism 16 moves currency bills with the
narrow dimension "W" of the bills being parallel to the
transport path and the scan direction. Thus, as a bill
17 moves on the transport path on the scanhead 18, the
coherent light strip 24 effectively scans the bill across
the narrow dimension "W" of the bill. Preferably, the
transport path is so arranged that a currency bill 17 is
scanned approximately about the central section of the
bill along its narrow dimension, as best shown in FIG. 1.
The scanhead 18 functions to detect light reflected
~ 91/11778 2 0 5 0 5 8 9 PCT/US91/00~3
from the bill as it moves across the illuminated light
strip 24 and to provide an analog representation of the
variation in light so reflected which, in turn, -
represents the variation in the "black" and "white"
content of the printed pattern or indicia on the surface
of the bill. This variation in light reflected from the
narrow dimension sc~nning of the bills serves as a
measure for disting~ h;ng~ with a high degree of
confidence, among a plurality of currency denominations
which the system of this invention is programmed to
handle.
A series of such detected reflectance signals are
obtained across the narrow dimension of the bill, or
across a selected segment thereof, and the resulting
analog signals are digitized under control of the CPU 30
to yield a fixed number of digital reflectance data
samples. The data samples are then subjected to a
- digitizing process which includes a normalizing routine
for processing the sampled data for improved correlation
and for smoothing out variations due to "contrast"
fluctuations in the printed pattern existing on the bill
surface. The normalized reflectance data so digitized
represents a characteristic pattern that is fairly unique
for a given bill denomination and provides sufficient
distinguishing features between characteristic patterns
for different currency denominations, as will be
explained in detail below.
In order to ensure strict correspondence between
reflectance samples obtained by narrow dimension scanning
30 of successive bills, the initiation of the reflectance
sampling process is preferably controlled through the CPU
30 ~y means of an optical encoder 32 which is linked to
the bill transport mechanism 16 and precisely tracks the
physical movement of the bill 17 across the scanhead 18.
More specifically, the optical encoder 32 is linked to
the rotary motion of the drive motor which generates the
movement imparted to the bill as it is relayed along the
~91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
transport path. In addition, it is ensured that positive
contact is maintained between the bill and the transport
path, particularly when the bill is being scanned by the
scanhead 18. Under these conditions, the optical encoder
is capable of precisely tracking the movement of the bill
relative to the light strip generated by the scanhead by
monitoring the rotary motion of the drive motor.
The ouL~ of photodetector 26 is monitored by the
CPU 30 to initially detect the presence of the bill
underneath the scanhead and, subsequently, to detect the
starting point of the printed pattern on the bill, as
represented by the thin borderline 17A which typically
encloses the printed indicia on currency bills. Once the
borderline 17A has been detected, the optical encoder is
used to control the timing and number of reflectance
samples that are obtained from the output of the
photodetector 26 as the bill 17 moves across the scanhead
18 and is scanned along its narrow dimension.
The detection of the borderline constitutes an
important step and realizes improved discrimination
efficiency since the borderline serves as an absolute
reference point for initiation of sampling. If the edge
of a bill were to be used as a reference point, relative
displacement of sampling points can occur because of the
random manner in which the distance from the edge to the
borderline varies from bill to bill due to the relatively
large range of tolerances permitted during printing a~d
cutting of currency bills. As a result, it becomes
difficult to establish direct correspondence between
sample points in successive bill scans and the
discrimination efficiency is adversely affected.
The use of the optical encoder for controlling the
sampling process relative to the physical movement of a
bill across the scanhead is also advantageous in that the
encoder can be used to provide a predetermined delay
following detection of the borderline prior to initiation
of samples. The encoder delay can be adjusted in such a
~91/11778 2 0 5 0 5 8 9 PCT/US91/00~3
12
way that the bill is scanned only across those segments
along its narrow dimension which contain the most
distinguishable printed indicia relative to the different
currency denominations.
In the case of U.S. currency, for instance, it has
been determined that the central, approximately two-inch
portion of currency bills, as scanned across the central
section of the narrow dimension of the bill, provides
sufficient data for distin~1;sh;ng between the various
U.S. currency denominations on the basis of the
correlation ~ech~;que of this invention. Accordingly,
the optical encoder can be used to control the sc~n~;ng
process so that reflectance samples are taken for a set
period of time and only after a certain period of time
has elapsed since the borderline has been detected,
thereby restricting the scanning to the desired central
portion of the narrow dimension of the bill.
The optical sensing and correlation technique is
based upon using the above process to generate a series
of master characteristic patterns using "new" or
"original" bills for each denomination of currency that
is to be detected. According to a preferred embodiment,
four characteristic patterns are generated and stored
within system memory, preferably in the form of an EEPROM
34 (see FIG. l), for each detectable currency
denomination. The characteristic patterns for each bill
are generated corresponding, respectively, to optical
scans, i.e., the process of obtaining the pre-determined
number of reflectance samples, performed on the "top" and
"bottom" surfaces of the bill taken both along the
"forward" and "reverse" directions relative to the
pattern printed on the bill.
In adapting the invented technique to U.S. currency,
for example, characteristic patterns are generated and
stored for seven different denominations of U.S.
currency, i.e., $1, $2, $~, $10, $20, $50 and $100.
Accordingly, a master set of 28 different characteristic
W~91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
13
patterns is stored within the system memory for
subsequent correlation purposes. Once the master
characteristic patterns have been stored, the pattern
generated by ~cAnning a bill under test is compared by
S the CPU 30 with each of the 28 pre-stored master
characteristic patterns to generate, for each comparison,
a correlation number representing the extent of
correlation, i.e., similarity between corresponding ones
of the plurality of data samples, for the patterns being
compared.
The CPU 30 is programmed to identify the
denomination of the scanned bill as corresponding to the
stored characteristic pattern for which the correlation
number resulting from pattern comparison is found to be
the highest. In order to preclude the possibility of
mischaracterizing the denomination of a scanned bill, as
well as to reduce the possibility of spurious notes being
identified as belonging to a valid denomination, a bi-
level threshold of correlation is used as the basis for
making a "positive" call, as will be explained in detail
below.
Using the above sensing and correlation approach,
the CPU 30 is programmed to count the number of bills
belonging to a particular currency denomination as part
of a given set of bills that have been scanned for a
given scan batch, and to determine the aggregate total of
the currency amount represented by the bills scanned
during a scan batch. The CPU 30 is also linked to an
ouL~uL unit 36 which is adapted to provide a display of
the number of bills counted, the breakdown of the bills
in terms of currency denomination, and the aggregate
total of the currency value represented by counted bills.
The output unit 36 can also be adapted to provide a
print-out of the displayed information in a desired
format.
Referring now to FIG. 2 there is shown a
representation, in block diagram form, of a preferred
' ` 91/11778 ~ 0 5 0 5 8 9 PCT/US91/~ ~3
14
circuit arrangement for processing and correlating
reflectance data according to the system of this
invention. As shown therein, the CPU 30 accepts and
processes a variety of input signals including those from
the optical encoder 32, the photodetector 26 and a memory
unit 38, which can be a static random access memory (RAM)
or an erasable programmable read only memory (EPROM).
The memory unit 38 has stored within it the correlation
program on the basis of which patterns are generated and
test patterns compared with stored master programs in
order to identify the denomination of test currency. A
crystal 40 serves as the time base for the CPU 30, which
is also provided with an external reference voltage VREF
on the basis of which peak detection of sensed
reflectance data is performed, as explained in detail
below.
The CPU 30 also accepts a timer reset signal from a
reset unit 44 which, as shown in FIG. 2A, accepts the
output voltage from the photodetector 26 and compares it,
by means of a threshold detector 44A, relative to a pre-
set voltage threshold, typically S.o volts, to provide a
reset signal which goes "high" when a reflectance value
corresponding to the presence of paper is sensed. More
specifically, reflectance sampling is based on the
premise that no portion of the illuminated light strip
(24 in FIG. 1) is reflected to the photodetector in the
absence of a bill positioned below the scanhead. Under
these conditions, the output of the photodetector
represents a "dark" or "zero" level reading. The
photodetector output changes to a "white" reading,
typically set to have a value of about 5.0 volts, when
the edge of a bill first becomes positioned below the
scanhead and falls under the light strip 24. When this
occurs, the reset unit 44 provides a "high" signal to the
CPU 30 and marks the initiation of the scanning
procedure.
WO9t/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
-
In accordance with a feature of this invention, the
thickness of the illuminated strip of light produced by
the light sources within the scanhead is set to be
relatively small for the initial stage of the scan when
the thin borderline is being detected. The use of the
narrow slit increases the sensitivity with which the
reflected light is detected and allows minute variations
in the "gray" level reflected off the bill surface to be
sensed. This is important in ensuring that the thin
borderline of the pattern, i.e., the starting point of
the printed pattern on the bill, is accurately detected.
Once the borderline has been detected, subsequent
reflectance sampling is performed on the basis of a
relatively wider light strip in order to completely scan
across the narrow dimension of the bill and obtain the
desired number of samples, at a rapid rate. The use of a
wider slit for the actual sampling also smoothens out the
o~L~L characteristics of the photodetector and realizes
the relatively large magnitude of analog voltage which is
essential for accurate representation and processing of
the detected reflectance values.
Returning to FIG. 2, the CPU 30 processes the output
of photodetector 26 through a peak detector 46 which
essentially functions to sample the photodetector output
voltage and hold the highest, i.e., peak, voltage value
encountered after the detector has been enabled. The
peak detector is also adapted to define a scaled voltage
on the basis of which the pattern borderline on bills is
detected. The detector 46 includes an ADC 48 for
digitizing the photodetector output and a digital-to-
analog convertor (DAC) 50 for reconverting the signals to
an analog form on the basis of the pre-selected reference
voltage VREF from the voltage source 42. The output of
DAC 50 is fed through an inverting amplifier 52 to a
voltage divider 54 which lowers the input voltage down to
a scaled voltage Vs representing a predefined percentage
of the peak value. The voltage Vs is based upon the
~09t/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
_
16
percentage drop in output voltage of the peak detector as
it reflects the transition from the "high" reflectance
value resulting from the sc~nn;ng of the unprinted edge
portions of a currency bill to the relatively lower
"gray" reflectance value resulting when the thin
borderline is encountered. Preferably, the scaled
voltage Vs is set to be about ~0 - 80 percent of the peak
voltage.
The scaled voltage Vs is supplied to a line detector
56 which is also provided with the incoming instantaneous
~L~uL of the photodetector 26. The line detector 56
compares the two voltages at its input side and generates
a signal ~T which normally stays "low" and goes "high"
when the edge of the bill is scanned. The signal ~ET
goes "low" when the incoming photodetector output reaches
the pre-defined percentage of the peak photodetector
o~uL up to that point, as represented by the voltage Vs.
Thus, when the signal ~ET goes "low", it is an indication
that the borderline of the bill pattern has been
detected. At this point, the CPU 30 initiates the actual
reflectance sampling under control of the encoder 32 (see
FIG. 2) and the desired fixed number of reflectance
samples are obtained as the currency bill moves across
the illuminated light strip and is scanned along the
-central section of its narrow dimension.
When master characteristic patterns are being
generated, the reflectance samples resulting from the
scanning of a "new" bill are loaded into corresponding
designated sections within a system memory 60, which is
preferably an EEPROM. The loading of samples is
accomplished through a buffered address latch 58, if
necessary. Preferably, master patterns are generated by
scanning a "new" bill a plurality of times, typically
three (3~ times, and obtaining the average of
corresponding data samples before storing the average as
representing a master pattern. During currency
discrimination, the reflectance values resulting from the
WO91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
_ 17
scAnning of a test bill are sequentially compared, under
control of the correlation program stored within the
-memory unit 38, with each of the corresponding
characteristic patterns stored within the EEPROM 60,
again through the address latch 58.
Referring now to FIGS. 3-7, there are shown flow
charts illustrating the sequence of operations involved
in implementing the above-described optical sensing and
correlation ~e~hnique of this invention. FIG. 3, in
particular, illustrates the sequence involved in
detecting the presence of a bill under the scanhead and
the borderline on the bill. This section of the system
program, designated as "TRIGGER", is initiated at step
70. At step 7l a determination is made as to whether or
not a start-of-note interrupt, which signifies that the
system is ready to search for a presence of a bill, is
set, i.e., has occurred. If the answer at step 71 is
found to be positive, step 72 is reached where the
presence of the bill below the scanhead is ascertained on
the basis of the reset procedure described above in
connection with the reset unit 44 of FIG. 2.
If the answer at step 72 is found to be positive,
i.e., a bill is found to be present, step 73 is reached
where a test is performed to see if the borderline has
been detected on the basis of the reduction in peak value
to a predefined percentage thereof, which, as described
above, is indicated by the signal ~ET going "low." If
the answer at step 73 is found to be negative, the
~o~am continues to loop until the borderline has been
detected. If the answer at step 72 is found to be
negative, i.e., no bill is found to be present, the
start-of-note interrupt is reset at step 74 and the
program returns from interrupt at step 75.
If the borderline is found to have been detected at
step 73, step 76 is accessed where an A/D completion
interrupt is enabled, thereby signifying that the analog-
to-digital conversion can subsequently be performed at
~vo 91/11778
2050589
18
desired time intervals. Next, at step 77, the time when
the first reflectance sample is to be obtained is
defined, in conjunction with the output of the optical
encoder. At step 78 the capture and digitization of the
detected reflectance samples is undertaken by recalling a
routine designated as "STARTA2D" which will be described
in detail below. At the completion of the digitization
process, the end-of-note interrupt is enabled at step 79,
which resets the system for sensing the presence of the
following bill to be scanned. Subsequently, at step 80
the program returns from interrupt. If the start-of-
note interrupt is not found to have occurred at step 71,
a determination is made at step 81 to see if the end-of-
note interrupt has occurred. If the answer at 81 is
negative, the program returns from interrupt at step 85.
If a positive answer is obtained at 81, step 83 is
accessed where the start-of-note interrupt is activated
and, at step 84, the reset unit, which monitors the
presence of a bill, is reset to be ready for determining
the presence of bills. Subsequently, the program returns
from interrupt at step 85.
Referring now to FIGS. 4A and 4B there are shown,
respectively, routines for starting the STARTA2D routine
and the digitizating routine itself. In FIG. 4A, the
initiation of the STARTA2D routine at step 90 causes the
sample pointer, which provides an indication of the
sample being obtained and digitized at a given time, to
be initialized. Subsequently, at step 91, the particular
channel on which the analog-to-digital conversion is to
be performed is enabled. The interrupt authorizing the
digitization of the first sample is enabled at step 92
and the main program accessed again at step 93.
FIG. 4B is a flow chart illustrating the sequential
procedure involved in the analog-to-digital conversion
routine, which is designated as "A2D". The routine is
started at step 100. Next, the sample pointer is
decremented at step 101 so as to maintain an indication
WO91/11778 2 0 5 0 5 8 9 PCT/US91~ ~3
19
of the number of samples r~r~;n;ng to be obtained. At
step 102, the digital data corresponding to the output of
the photodetector for the current sample is read. The
data is converted to its final form at step 103 and
stored within a pre-defined memory segment as Xl~.
Next, at step 105, a check is made to see if the
desired fixed number of samples "N" has been taken. If
the answer is found to be negative, step 106 is accessed
where the interrupt authorizing the digitization of the
s~cce~ing sample is enabled and the ~o~,am returns from
interrupt at step 107 for completing the rest of the
digitizing process. However, if the answer at step 105
is found to be positive, i.e., the desired number of
samples have already been obtained, a flag indicating the
same is set at step 108 and the program returns from
interrupt at step 109.
Referring now to FIG. 5, there is shown the
sequential procedure involved in executing the routine,
designated as "EXEC", which performs the mathematical
steps involved in the correlation process. The routine
is started at step 110. At step 111, all interrupts are
disabled while CPU initialization occurs. At step 112,
any constants associated with the sampling process are
set and, at step 113, communications protocols, if any,
for exchange of processed data and associated results,
bad rates, interrupt masks, etc. are defined.
At step 114, the reset unit indicating the presence
of a bill is reset for detecting the presence of the
first bill to be scanned. At step 115, the start-of-note
interrupt is enabled to put the system on the look out
for the first incoming bill. Subsequently, at step 116,
all other related interrupts are also enabled since, at
this point, the initialization process has been completed
and the system is ready to begin scanning bills. A check
is made at step 117 to see if, in fact, all the desired
number of samples have been obtained. If the answer at
step 117 is found to be negative the program loops until
WO91/11778 2 0 5 0 5 8 9 PCT/US9t/00~3
a positive answer is obtained. At that time, step 118 is
accessed where a flag is set to indicate the initiation
of the correlation procedure.
In accordance with this invention, a simple
correlation procedure is utilized for processing
digitized reflectance values into a form which is
conveniently and accurately compared to corresponding
values pre-stored in a identical format. More
lo specifically, as a first step, the
mean value X for the set of digitized reflectance samples
(comparing "n" samples) obtained for a bill scan run is
first obtained as below:
n Xj
X = ~ _ (1)
i=0 n
Subsequently, a normalizing factor Sigma "a" is
determined as being equivalent to the sum of the square
of the difference between each sample and the mean, as
normalized by the total number of samples. More
specifically, the normalizing factor is calculated as
below:
n IXj-XI2
a = ~ . .(2)
i=0 n
In the final step, each reflectance sample is
normalized by obtaining the difference between the sample
and the above-calculated mean value and dividing it b~
the square root of the normalizing factor Sigma "a" as
defined by the following equation:
Xj--X
Xn = . . . .(3)
(a,~
The result of using the above correlation equations
is that, subsequent to the normalizing process, a
relationship of correlation exists between a test pattern
and a master pattern such that the aggregate sum of the
products of corresponding samples in a test pattern and
any master pattern, when divided by the total number of
WO91/11778 - 2 0 5 0 5 89 PCT/US91/002B3
21
samples, equals unity if the patterns are identical.
Otherwise, a value less than unity is obtained.
Accordingly, the correlation number or factor resulting
from the comparison of normalized samples within a test~
pattern to those of a stored master pattern provides a
clear indication of the degree of similarity or
correlation between the two patterns.
According to a preferred embodiment of this
invention, the fixed number of reflectance samples which
are digitized and normalized for a bill scan is selected
to be 64. It has experimentally been found that the use
of higher binary orders of samples (such as 128, 256,
etc.) does not provide a correspondingly increased
discrimination efficiency relative to the increased
processing time involved in implementing the above-
described correlation procedure. It has also been found
that the use of a binary order of samples lower than 64,
such as 32, produces a substantial drop in discrimination
efficiency.
The correlation factor can be represented
conveniently in binary terms for ease of correlation. In
a preferred embodiment, for instance, the factor of unity
which results when hundred percent correlation exists is
represented in terms of the binary number 210, which is
equal to a decimal value of 1024. Using the aboveprocedure, the normalized samples within a test pattern
are compared to each of the 28 master characteristic
patterns stored within the system memory in order to
determine the particular stored pattern to which the test
pattern most corresponds by identifying the comparison
which yields a correlation number closest to 1024.
According to a feature of this invention, a bi-level
threshold of correlation is required to be satisfied
before a particular call is made. More specifically, the
correlation procedure is adapted to identify the two
highest correlation numbers resulting from the comparison
of the test pattern to one of the stored patterns. At
W091/11778 PCT/US91/~ ~3
2050589
22
that point, a minimum threshold of correlation is
required to be satisfied by these two correlation
numbers. It has experimentally been found that a
correlation number of about 800 serves as a good cut-off
threshold above which positive calls may be made with a
high degree of confidence and below which the designation
of a test pattern as corresponding to any of the stored
patterns is uncertain. As-a second thresholding level, a
minimum separation is prescribed between the two highest
correlation numbers before making a call. This ensures
that a positive call is made only when a test pattern
does not correspond, within a given range of correlation,
to more than one stored master pattern. Preferably, the
minimum separation between correlation numbers is set to
be between 100-150.
Returning now to FIG. 5, the correlation procedure
is initiated at step 119 where a routine designated as
"PROCESS" is accessed. The procedure involved in
executing this routine is illustrated at FIG. 6A which
shows the routine starting at step 130. At step 131, the
mean X is calculated on the basis of Equation (1). At
step 132 the sum of the squares is calculated in
accordance with Equation (2). At step 133, the digitized
values of the reflectance samples, as represented in
integer format, are converted to floating point format
for further processing. At step 134, a check is made to
see if all samples have been processed and if the answer
is found to be positive, the routine ends at step 135 and
the main program is accessed again. If the answer at
step 134 is found to be negative, the routine returns to
step 132 where the above calculations are repeated.
At the end of the routine P~OCESS, the program
returns to the routine EXEC at step 120 where the flag
indicating that all digitized reflectance samples have
been processed is reset. Subsequently, at step 121, a
routine designated as "SIGCAL" is accessed. The
procedure involved in executing this routine is
WO91/11778 2 0 5 0 5 8 9 PCT/US91/00~3
- 23
illustrated at FIG. 6B which shows the routine starting
at step 140. At step 141, the square root of the sum of
the squares, as calculated by the routine PROC~SS, is
calculated in accordance with Equation (2). At step 142,
the floating point values calculated by the routine
PROCESS are normalized in accordance with Equation (3)
using the calculated values at step 141. At step 143, a
check is made to see if all digital samples have been
processed. If the answer at step 143 is found to be
negative, the program returns to step 142 and the
conversion is continued until all samples have been
processed. At that point, the answer at step 143 is
positive and the routine returns to the main program at
step 144.
Returning to the flow chart of FIG. 5, the next step
to be executed is step 12`2 where a routine designated as
"CORREL" is accessed. The procedure involved in
executing this routine is illustrated at FIG. 7 which
shows the routine starting at 150. At step 151,
correlation results are initialized to zero and, at step
152, the test pattern is compared to the first one of the
stored master patterns. At step 153, the first call
corresponding to the highest correlation num~er obtained
up to that point is determined. At step 154, the second
call corresponding to the second highest correlation
number obtained up to that point is determined. At step
155, a check is made to see if the test pattern has been
compared to all master patterns. If the answer is found
to be negative, the routine reverts to step 152 where the
comparison procedure is reiterated. When all master
patterns have been compared to the test pattern, step 155
yields a positive result and the routine returns to the
main program at step 156.
Returning again to FIG. 5, at step 123, a flag
indicating that the correlation procedure has been
completed is reset and step 124 is accessed where a
routine designated as "SEROUT" is initiated. It should
"'O91/11778 2 0 5 0 5 8 9 PCT/US9l/~ ~3
-24
be noted that steps 118 and 123, which are directed to
the setting and resetting of the flag TP2, primarily
function to provide a measure of the processing time
involved in the overall correlation procedure. These
steps can be dispensed with, if processing time is not
being monitored. The procedure involved in executing the
routine SEROUT is illustrated at FIG. 8 which shows the
routine as starting at step 160. At step 161, the
currency denomination corresponding to the first call is
converted to ASCII format and displayed. At step 162,
the correlation number corresponding to the first call is
converted to ASCII format and displayed.
At step 163, the currency denomination corresponding
to the second call is converted to ASCII format and
displayed. At step 164, the correlation number
corresponding to the second call is converted to ASCII
format and displayed. Subsequently, the routine returns
to the main program. At this point in the main program,
the correlation procedure is completed and any related
counting of identified denominations may be executed with
the associated results also being displayed along with
the corresponding calls and correlation numbers. After
this correlation and display procedure is completed, the
system is ready for initiating the process of scanning
the next incoming currency bill.
It should be noted that, in implementing the optical
sensing and correlation technique of this invention,
separate microprocessor units may be provided for (i)
accomplishing the sampling and correlation process and
(ii) for controlling the general functions of the overall
system. In such an implementation, the general processor
unit would preferably be used for displaying the
identified denominations and any associated counting
results. With this approach, the routine SEROUT (FIG. 8)
would merely involve the transmission of the bill
denomination and call information from the sampling and
WO91/11778 PCT/US91/~ ~3
2050589
_ 25
correlation processor unit to the general processor unit
for subsequent display.
Referring now to FIGS. 9A-C there are shown three
test patterns generated, respectively, for the forward
scanning of a $1 bill along its top face, the reverse
Cc~nning of a ~2 bill on its top face, and the forward
scanning of a $100 bill about its top face. It should be
noted that, for ~u~o~es of clarity the test patterns in
FIGS. 9A-C were generated by using 128 reflectance
samples per bill scan, as opposed to the preferred use of
only 64 samples. The marked difference existing between
corresponding samples for these three test patterns is
indicative of the high degree of confidence with which
currency denominations may be called using the foregoing
optical sensing and correlation procedure.
The optical sensing and correlation technique
described above permits identification of pre-programmed
currency denominations with a high degree of accuracy and
is based upon a relatively low processing time for
digitizing sampled reflectance values and comparing them
to the master characteristic patterns. The approach is
used to scan currency bills, normalize the scanned data
and generate master patterns in such a way that bill
scans during operation have a direct correspondence
between compared sample points in portions of the bills
which possess the most distinguishable printed indicia.
A relatively low number of reflectance samples is
required in order to be able to adequately distinguish
between several currency denominations.
A major advantage with this approach is that it is
not required that currency bills be scanned along their
wide dimensions. Conventional systems have been forced
to adopt the wide dimension scanning approach in order to
obtain the larger number of samples typically required
for accurate denomination discrimination. Further, the
reduction in the number of samples reduces the processing
time to such an extent that additional comparisons can be
~091/11778 2 ~ 5 0 5 8 9 PCT/US91/~283
26
made during the time available between the scanning of
successive bills. More specifically, as described above,
it becomes possible to compare a test pattern with at
least four stored master characteristic patterns so that
the system is made capable of identifying currency which
is scanned in the "forward" or "reverse" directions along
the "top" or "bottom" surfaces of bills.
Another advantage accruing from the reduction in
processing time realized by the present sensing and
correlation scheme is that the response time involved in
either stopping the transport of a bill that has been
identified as "spurious", i.e., not corresponding to any
of the stored master characteristic patterns, or
diverting such a ~ill to a separate stacker bin, is
correspondingly shortened. Accordingly, the system can
conveniently be programmed to set a flag when a scanned
pattern does not correspond to any of the master
patterns. The identification of such a condition can be
used to stop the drive bill transport motor for the
mechanism. Since the optical encoder is tied to the
rotational movement of the drive motor, synchronism can
be maintained between pre- and post-stop conditions. In
the dual-processor implementation discussed above, the
information concerning the identification of a "spurious"
bill would be included in the information relayed to the
general processor unit which, in turn, would control the
drive motor appropriately.
The correlation procedure and the accuracy with
which a denomination is identified directly relates to
the degree of correspondence between reflectance samples
on the test pattern and corresponding samples on the
stored master patterns. Thus, shrinkage of "used" bills
which, in turn, causes a corresponding reduction in their
narrow dimension, can possibly produce a drop in the
degree of correlation between such used bills of a given
denomination and the corresponding master patterns.
Currency bills which have experienced a high degree of
WO91/11778 PCT/US91/~ ~3
2050589
27
usage exhibit such a reduction in both the narrow and
wide ~;~Gnsions of the bills. While the sensing and
correlation technique of this invention remains
relatively independent of any changes in the wide
S dimension of bills, reduction along the narrow dimension
can affect correlation factors by realizing a relative
displacement of reflectance samples obtained as the
"shrunk" ~ills are transported across the scanhead.
In order to accommodate or nullify the effect of
such narrow dimension shrinking, the above-described
correlation te~h~;que can be modified by use of a
progressive shifting approach whereby a test pattern
which does not correspond to any of the master patterns
is partitioned into predefined sections, and samples in
successive sections are progressively shifted and
compared again to the stored patterns in order to
identify the denomination. It has experimentally been
determined that such progressive shifting effectively
counteracts any sample displacement resulting from
shrinkage of a bill along its narrow dimension.
The progressive shifting effect is best illustrated
by the correlation patterns shown in FIGS. lOA-D. For
purposes of clarity, the illustrated patterns were
generated using 128 samples for each bill scan as
compared to the preferred use of 64 samples. FIG. lOA
shows the correlation between a test pattern (represented
by a heavy line) and a corresponding master pattern
(represented by a thin line). It is clear from FIG. lOA
that the degree of correlation between the two patterns
is relatively low and exhibits a correlation factor of
606.
The manner in which the correlation between these
patterns is increased by employing progressive shifting
is best illustrated by considering the correlation at the
reference points designated as A-E along the axis
defining the number of samples. The effect on
correlation produced by "single" progressive shifting is
~0 91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
-
28
shown in FIG. 10B which shows "single~ shifting of the
test pattern of FIG. 10A. This is affected by dividing
the test pattern into two equal segments each comprising
64 samples. The first segment is retained without any
S shift, whereas the second segment is shifted by a factor
of one data sample. Under these conditions, it is found
that the correlation factor at the reference points
located in the shifted section, particularly at point E,
is improved.
FIG. 10C shows the effect produced by "double"
progressive shifting whereby sections of the test pattern
are shifted in three stages. This is accomplished by
dividing the overall pattern into three approximately
egual sized sections. Section one is not shifted,
section two is shifted by one data sample (as in FIG.
10B), and section three is shifted by a factor of two
data samples. With "double" shifting, it can be seen
that the correlation factor at point E is further
increased.
On a similar basis, FIG. 10D shows the effect on
correlation produced by "triple" progressive shifting
where the overall pattern is first divided into four (4)
approximately equal sized sections. Subsequently,
section one is retained without any shift, section two is
shifted by one data sample, section three is shifted by
two data samples, and section four is shifted by three
data samples. Under these conditions, the correlation
factor at point E is seen to have increased again.
FIG. 10E shows the effect on correlation produced by
"quadruple" shifting, where the pattern is first divided
into five (5) approximately equal sized sections. The
first four (4) sections are shifted in accordance with
the "triple" shifting approach of FIG. 10D, whereas the
fifth section is shifted by a factor of four (4) data
samples. From FIG. 10E it is clear that the correlation
at point E is increased almost to the point of
superimposition of the compared data samples.
WO9l/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
_ 29
The advantage of using the progressive shifting
approach, as opposed to merely shifting by a set amount
of data samples across the overall test pattern, is that
the improvement in correlation achieved in the initial
sections of the pattern as a result of shifting is not
neutralized or offset by any subsequent shifts in the
test pattern. It is apparent from the above figures that
the degree of correlation for sample points falling
within the p~o~essively.shifted sections increases
correspondingly.
More importantly, the progressive shifting realizes
substantial increases in the overall correlation factor
resulting from pattern comparison. For instance, the
original correlation factor of 606 (FIG. lOA) is
increased to 681 by the "single" shifting shown in FIG.
lOB. The "double" shifting shown in FIG. lOC increases
the correlation number to 793, the "triple" shifting of
FIG. lOD increases the correlation number to 906, and,
finally, the "quadruple" shifting shown in FIG. lOE
increases the overall correlation number to 960. Using
the above approach, it has been determined that used
currency bills which exhibit a high degree of narrow
dimension shrin~age and which cannot be accurately
identified as belonging to the correct currency
denomination when the correlation is performed without
any shifting, can be identified with a high degree of
certainty by using progressive shifting approach,
preferably by adopting "triple" or "quadruple" shifting.
Referring now to FIG. 11, there is shown apparatus
210 for currency discrimination and counting which
embodies the principles of the present invention. The
apparatus comprises a housing 212 which includes left and
right hand sidewalls 214 and 216, respectively, a rear
wall 218, and a top surface generally designated as 220.
The apparatus has a front section 222 which comprises a
generally vertical forward section 224 and a forward
sloping section 225 which includes side sections provided
'~0 91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
.
with control panels 226A and 226B upon which various
control switches for operating the apparatus, as well as
associated display means, are mounted.
For accepting a stack of currency bills 228 which
have to be discriminated according to denomination, an
input bin 227 is defined on the top surface 220 by a
downwardly sloping support surface 229 on which are
provided a pair of vertically disposed side walls 230,
232 linked together by a vertically disposed front wall
234. The walls 230, 232 and 234, in combination with the
sloping surface 229, define an enclosure where the stack
of currency bills 228 is positioned.
From the input bin, currency bills are moved along a
tri-sectional transport path which includes an input path
where bills are moved along a first direction in a
substantially flat position, a curved guideway where
bills are accepted from the input path and guided in such
a way as to change the direction of travel to a second
different direction, and an output path where the bills
20 are moved in a flat position along the second different
direction across currency discrimination means located
downstream of the curved guideway, as will be described
in detail below. In accordance with the improved optical
sensing and correlation technique of this invention, the
25 transport path is defined in such a way that currency
bills are accepted, transported along the input path, the
curved guideway, and the output path, and stacked with
the narrow dimension "W" of the bills being maintained
parallel to the transport path and the direction of
3 0 movement at all times.
The forward sloping section 225 of the document
handling apparatus 210 includes a platform surface 235
centrally disposed between the side walls 214, 216 and is
adapted to accept currency bills which have been
35 processed through the currency discrimination means for
being delivered to a stacker plate 242 where the
processed bills are stacked for subsequent removal. More
WO91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
specifically, the platform 235 includes an associated
angular surface 236 and is provided with openings 237,
237A from which flexible blades 238A, 240A of a
corresponding pair of stacker wheels 238, 240,
respectively, extend outwardly. The stacker wheels are
supported for rotational movement about a stacker shaft
241 disposed about the angular surface 236 and suspended
across the side walls 214 and 216. The flexible
blades 238A, 240A of the stacker wheels cooperate with
the stacker platform 235 and the openings 237, 237A to
pick up currency bills delivered thereto. The blades
operate to subsequently deliver such bills to a stacker
plate 242 which is linked to the angular surface 236 and
which also accommodates the stacker wheel openings and
the wheels projecting therefrom. During operation, a
currency bill which is delivered to the stacker platform
235 is picked up by the flexible blades and becomes
lodged between a pair of adjacent blades which, in
combination, define a curved enclosure which decelerates
a bill entering therein and serves as a means for
supporting and transferring the bill from the stacker
platform 235 onto the stacker plate 242 as the stacker
wheels rotate. The mechanical configuration of the
stacker wheels and the flexible blades provided
thereupon, as well as the manner in which they cooperate
with the stacker platform and the stacker plate, is
conventional and, accordingly, is not described in detail
herein.
The bill handling and counting apparatus 210 is
provided with means for picking up or "stripping"
currency bills, one at a time, from bills that are
stacked in the input bin 227. In order to provide this
stripping action, a feed roller 246 is rotationally
suspended about a drive shaft 247 which, in turn, is
supported across the side walls 214, 216. The feed
roller 246 projects through a slot provided on the
downwardly sloping surface 229 of the input bin 227 which
'-- ' 91/11778 2 0 5 0 5 8 9 PCT/US91/00283
-
32
defines the input path and is in the form of an eccentric
roller at least a part of the periphery of which is
provided with a relatively high friction-bearing surface
246A. The surface 246A is adapted to engage the bottom
bill of the bill stack 228 as the roller 246 rotates;
this initiates the advancement of the bottom bill along
the feed direction represented by the arrow 247B (see
FIG. 13). The eccentric surface of the feed roller 246
essentially "jogs" the bill stack once per revolution so
as to agitate and loosen the bottom currency bill within
the stack, thereby facilitating the advancement of the
bottom bill along the feed direction.
The action of the feed roller 246 is supplemented by
the provision of a capstan or drum 248 which is suspended
for rotational movement about a capstan drive shaft 249
which, in turn, is supported across the side walls 214
and 216. Preferably, the capstan 248 comprises a
centrally disposed friction roller 248A having a smooth
surface and formed of a friction-bearing material such as
rubber or hard plastic. The friction roller is
sandwiched between a pair of capstan rollers 248B and
248C at least a part of the external periphery of which
are provided with a high friction-bearing surface 248D.
The friction surface 248D is akin to the friction
surface 246A provided on the feed roller and permits the
capstan rollers to frictionally advance the bottom bill
along the feed direction. Preferably, the rotational
movement of the capstan 248 and the feed roller 246 is
synchronized in such a way that the frictional surfaces
provided on the peripheries of the capstan and the feed
roller rotate in unison, thereby inducing complimentary
frictional contact with the bottom bill of the bill stac~
228.
In order to ensure active contact between the
capstan 248 and a currency bill which is jogged by the
feed roller 246 and is in the process of being advanced
WO91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
33
picker rollers 252A, 252B, are provided for exerting a
consistent downward force onto the leading edges of the
currency bills stationed in the input bin 227. The
picker rollers are supported on corresponding picker arms
254A, 254B which, in turn, are supported for arcuate
movement about a support shaft 256 suspended across the
side walls of the apparatus. The picker rollers are free
wheeling about the picker arms and when there are no
currency bills in contact with the capstan 248, bear down
upon the friction roller 248A and, accordingly, are
induced into counter-rotation therewith. However, when
currency bills are present and are in contact with the
capstan 248, the picker rollers bear down into contact
with the leading edges of the currency bills and exert a
direct downward force on the bills since the rotational
movement of rollers is inhibited. The result is that the
advancing action brought about by contact between the
friction-bearing surfaces 248D on the capstan rollers
248B, 248C is accentuated,-thereby facilitating the
stripping away of a single currency bill at a time from
the bill stack 228.
In between the picker arms 254A, 254B, the support
shaft 256 also supports a separator arm 260 which carries
at its end remote from the shaft a stationary stripper
shoe 258 which is provided with a frictional surface
which imparts a frictional drag upon bills onto which the
picker rollers bear down. The separator arm is mounted
for arcuate movement about the support shaft 256 and is
spring loaded in such a way as to bear down with a
selected amount of force onto the capstan.
In operation, the picker rollers rotate with the
rotational movement of the friction roller 248A due to
their free wheeling nature until the leading edges of one
or more currency bills are encountered. At that point,
the rotational movement of the picker rollers stops and
the leading edges of the bills are forced into positive
contact with the friction bearing surfaces on the
~'~91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
34
periphery of the capstan rollers. The effect is to force
the bottom bill away from the rest of the bills along the
direction of rotation of the capstan. At the same time,
the separator shoe 258 also bears down on any of the
bills that are propelled forward by the capstan rollers.
The tension on the picker arm 254A is selected to be
such that the downward force exerted upon such a
propelled bill allows only a single bill to move
forward. If two or more bills happen to be propelled out
of the contact established between the picker rollers and
the capstan rollers, the downward force exerted by the
spring loaded shoe should be sufficient to inhibit
further forward movement of the bills. The tension under
which the picker arm is spring loaded can be conveniently
adjusted to control the downward bearing force exerted by
the shoe in such a way as to compliment the bill
stripping action produced by the picker rollers and the
capstan rollers. Thus, the possibility that more than
two bills may be propelled forward at the same time due
to the rotational movement of the capstan is
significantly reduced.
The bill transport path includes a curved guideway
270 provided in front of the capstan 248 for accepting
currency bills that have been propelled forward along the
input path defined by the forward section of the sloping
surface 229 into frictional contact with the rotating
capstan. The guideway 270 includes a curved section 272
which corresponds substantially to the curved periphery
of the capstan 248 so as to compliment the impetus
provided by the capstan rollers 248B, 248C to a stripped
currency bill.
A pair of idler rollers 262A, 262B is provided
downstream of the picker rollers for guiding bills
propelled by the capstan 248 into the curved guideway
270. More specifically, the idler rollers are mounted on
corresponding idler arms 264A, 264B which are mounted for
WO91/11778 ~ PCT/US91/00~3
2050589
arcuate movement about an idler shaft 266 which, in turn,
is supported across the side walls of the apparatus. The
idler arms are spring loaded on the idler shaft so that a
selected downward force can be exerted through the idler
rollers onto a stripped bill, thereby ensuring continued
contact between the bill and the capstan 248 until the
bill is guided into the curved section 272 of the
guideway 270.
Downstream of the curved section 272, the bill
transport path has an o~L~u~ path for currency bills.
The ouL~u~ path is provided in the form of a flat section
274 along which bills which have been guided along the
curved guideway 270 by the idler rollers 262A, 262B are
moved along a direction which is opposite to the
direction along which bills are moved out of the input
bin. The movement of bills along the direction of
rotation of the capstan, as induced by the picker rollers
252A, 252B and the capstan rollers 248B, 248C, and the
guidance provided by the section 272 of the curved
guideway 270 changes the direction of movement of the
currency bills from the initial movement along the
sloping surface 229 of input bin 227 (see arrow 247B in
FIG. 13) to a direction along the flat section 274 of the
output path, as best illustrated in FIG. 13 by the arrow
272B.
Thus, a currency bill which is stripped from the
bill stack in the input bin is initially moved along the
input path under positive contact between the picker
rollers 252A, 252B and the capstan rollers 248B, 248C.
Subsequently, the bill is guided through the curved
guideway 270 under positive contact with the idler
rollers 262A, 262B onto the flat section 274 of the
output path.
In the output path, currency bills are positively
guided along the flat section 274 by means of a transport
roller arrangement which includes a plurality of axially
spaced, positively driven transport rollers 282A, 284A,
V-~91/11778 2 0 5 0 5 8 9 PCT/US91/00~3
36
286A which are disposed on a transport shaft 287
supported across the side walls of the apparatus. The
flat section is provided with openings through which at
least two of the transport rollers, specifically rollers
S 282A and 284A, project into counter-rotating contact with
corresponding freewheeling passive rollers 292A, 294A.
The passive rollers are mounted on a support shaft 295
supported between the side walls of the apparatus below
the flat section 274 of the u~L~u~ path. The passive
transport rollers 292A, 294A are spring-loaded into
counter-rotating contact with the active transport
rollers 282A, 284A, 286A and the points of contact are
made coplanar with the output path so that currency bills
can be moved along the path in a flat manner under the
positive contact of the opposingly disposed active and
passive rollers. A similar set of active transport
rollers 282B, 284B, 286B and opposing spring-loaded
passive transport rollers 292B, 294B are provided
downstream of the first set of transport rollers at a
distance which is just short of the length of the narrow
dimension of the currency bills that are to be
discriminated. ~urther, the distance between the idler
rollers 262A, 262B and the first set of transport rollers
is selected to be such that a currency bill which is
guided along the curved guideway 259 is pulled into
contact between the first set of active and passive
transport rollers just before the bill moves away from
the positive contact between the idler rollers 262A, 262B
and the capstan 248.
The active transport rollers are driven at a speed
substantially higher than that of the capstan rollers.
Since the passive rollers are freewheeling and the active
rollers are positively driven, the first set of transport
rollers cause a bill that comes in between the rollers
3S along the flat section of the output path to be pulled
into the nip formed between the active and passive
rollers. The higher speed of the active transport
WO91/11778 2 0 5 0 5 8 9 PCT/US91/~ ~3
-
37
rollers imparts an abrupt acceleration to the bill; this
acceleration functions to separate or strip the bill away
from any other bills that may have been guided into the
curved guideway along with the bill being acted upon by
the transport rollers.
Downstream of the first set of transport rollers,
currency bills are moved along the flat section into the
nip formed between the second set of active and passive
transport rollers, which are driven at the same speed as
that of the first set of transport rollers. Preferably,
the opposing sets of active transport rollers 282A-282B,
284A-284B, and 286A-286B are linked together by a belt
290 so that the positive rotating action of the transport
shaft 287 is imparted to the rollers carried on the
second transport shaft 288. The disposition of the
second set of transport rollers is such that the positive
contact exerted by the rollers on a currency bill moving
along the output path occurs before the bill is released
from the positive contact between the first set of
transport rollers. The second set of transport rollers,
thus, positively guides a currency bill onto the stacker
platform 235 from where the stacker wheels 238, 240 pick
up the bill and deposit it onto the stacker plate 242.
Referring now in particular to FIGS. 14 and 15,
there are shown side and top views, respectively, of the
document processing apparatus of FIGS. 11-13, which
illustrate the mechanical arrangement for driving the
various means for transporting currency bills along the
three sections of the transport path, i.e., along the
input path, the curved guideway and the output path. As
shown therein, a motor 300 is used to impart rotational
movement to the capstan shaft 249 by means of a
belt/pulley arrangement comprising a pulley 310 provided
on the capstan shaft 249 and which is linked to a pulley
304 provided on the motor drive shaft through a belt 306.
The diameter of the driver pulley 310 is selected to be
appropriately larger than that of the motor pulley 304 in
W~t/11778 2 0 5 0 5 8 ~ PCT/US91/~ ~3
38
order to achieve the desired speed reduction from the
typically high speed at which the motor 300 operates.
The drive shaft 247 for the drive roller 246 is
provided with rotary motion by means of a pulley 308
provided thereupon which is linked to a corresponding
pulley 310 provided on the capstan shaft 249 through a
belt 312. The pulleys 308 and 310 are of the same
diameter so that the drive roller shaft 247 and, hence,
the drive roller 246, rotate in unison with the capstan
248 mounted on the capstan shaft 249.
In order to impart rotational movement to the
transport rollers, a pulley 314 is mounted on the
transport roller shaft 287 corresponding to the first set
of transport rollers and is linked to a corresponding
pulley 316 on the capstan shaft 249 through a belt 318.
The diameter of the transport roller pulley 314 is
selected to be appropriately smaller than that of the
corresponding capstan pulley 316 so as to realize a
stepping-up in speed from the capstan rollers to the
transport rollers. The second set of transport rollers
mounted on the transport roller shaft 288 is driven at
the same speed as the rollers on the first set of
transport rollers by means of a pulley 320 which is
linked to the transport pulley 314 by means of a belt
322.
As also shown in FIGS. 14 and 15, an optical encoder
299 is mounted on one of the transport roller shafts, ~
preferably the passively driven transport shaft 288, for
precisely tracking the lateral displacement of bills
supported by the transport rollers in terms of the
rotational movement of the transport shafts, as discussed
in detail above in connection with the optical sensing
and correlation technique of this invention.
In order to drive the stacker wheels 238, 240 an
intermediate pulley 322 is mounted on suitable support
means (not shown) and is linked to a corresponding pulley
324 provided on the capstan shaft 249 through a belt 326.
WO91/11778 2 0 5 0 5 8 ~ PCT/US91/~ ~3
39
Because of the time required for transporting currency
bills which have been stripped from the currency stack in
the input bin through the tri-sectional transport path
and onto the stacker platform, the speed at which the
stacker wheels can rotate for delivering processed bills
to the stacker plate is nececc~rily less than that of the
capstan shaft. Accordingly, the diameter of the
intermediate pulley 322 is-selected to be larger than
that of the correspon~;ng capstan pulley 324 so as to
realize a reduction in speed. The interim pulley 322 has
an associated pulley 328 which is linked to a stacker
pulley 330 provided on the drive shaft 241 for the
stacker wheels 238, 240 by means of a belt 332. In the
preferred embodiment shown in FIGS. ll-l5, the stacker
wheels 238, 240 rotate in the same direction as the
capstan rollers. This is accomplished by arranging the
belt 332 between the pulleys 328, 330 in a "Figure-8"
configuration about an anchoring pin 333 disposed between
the two pulleys.
The curved section 272 of the guideway 270 is
provided on its underside with an optical sensor
arrangement 299, including an LED 298, for performing
s~n~Ard currency handling operations such as counterfeit
detection using conventional techniques, doubles
detection, length detection, skew detection, etc.
However, unlike conventional arrangements, currency
discrimination according to denomination is not perfo~med
in this area, for reasons described below.
According to a feature of this invention, optical
scanning of currency bills, in accordance with the above-
described improved optical sensing and correlation
technique, is performed by means of an optical scanhead
296 which is disposed downstream of the curved guideway
270 along the flat section 274 of the output path. More
specifically, the scanhead 296 is located under the flat
section of the output path between the two sets of
transport rollers. The advantage of this approach is
'"O 91/11778 2 0 5 0 5 8 9 PCT/US9t/~ ~3
that optical scanning is performed on bills when they are
maintained in a substantially flat position as a result
of positive contact between the two sets of transport
rollers at both ends of the bill along their narrow
dimension.
It should be understood that the above-described
drive arrangement is provided for illustrative purposes
only. Alternate arrangements for imparting the necP~s~ry
rotational movement to generate movement of currency
bills along the tri-sectional transport path can be used
just as effectively. It is important, however, that the
surface speed of currency bills across the two sets of
transport rollers be greater than the surface speed of
the bills across the capstan rollers in order to achieve
optimum bill separation. It is this difference in speed
that generates the abrupt acceleration of currency bills
as the bills come into contact with the first set of
transport rollers.
The drive arrangement may also include a one-way
clutch (not shown) provided on the capstan shaft and the
capstan shafts, the transport roller shafts and the
stacker wheel shafts may be fitted with fly-wheel
arrangements (not shown). The combination of the one-way
clutch and the fly wheels can be used to advantage in
accelerated batch processing of currency bills by
ensuring that any bills remaining in the transport path
after currency discrimination are automatically pulled
off the transport path into the stacker plate as a result
of the inertial dynamics of the fly wheel arrangements.
As described above, implementation of the optical
sensing and correlation technique of this invention
requires only a relatively low number of reflectance
samples in order to adequately distinguish between
several currency denominations. Thus, highly accurate
discrimination becomes possible even though currency
bills are scanned along their narrow dimension. However,
the accuracy with which a denomination is identified is
WO 91/11778
2050589
-
41
based on the degree of correlation between reflectance
samples on the test pattern and corresponding samples on
the stored master patterns. Accordingly, it is important
that currency bills be transported across the
discrimination means in a flat position and, more
importantly, at a uniform speed.
This is achieved in the bill handling apparatus of
FIGS. 11-15, by positioning the optical scanhead 296 on
one side of the flat section 274 of the ouL~uL path
between the two sets of transport rollers. In this area,
currency bills are maintained in positive contact with
the two sets of rollers, thereby ensuring that the bills
move across the scanhead in a substantially flat fashion.
Further, a uniform speed of bill movement is maintained
in this area because the second set of passive transport
rollers is driven at a speed identical to that of the
active transport rollers by means of the belt connecting
the two sets of rollers. Disposing the optical scanhead
296 in such a fashion downstream of the curved guideway
270 along the flat section 274 maintains a direct
correspondence between reflectance samples obtained by
the optically scanning of bills to be discriminated and
the corresponding samples in the stored master patterns.
According to a preferred embodiment, the optical
scanhead comprises a plurality of light sources acting in
combination to uniformly illuminate light strips of the
desired dimension upon curxency bills positioned on the
transport path below the scanhead. As illustrated in
FIG. 16, the scanhead 296 includes a pair of LEDs 340,
342, directing beams of light 340A and 340B,
respectively, downwardly onto the flat section 274 of the
output path against which the scanhead is positioned.
The LEDs 340, 342 are angularly disposed relative to the
vertical axis Y in such a way that their respective light
beams combine to illuminate the desired light strip 342.
' ~9l/11778 2 0 5 0 5 8 9 PCT/US91/~283
42
The scanhead 296 includes a photodetector 346
centrally disposed directly above the strip for sensing
the light reflected off the strip. The photodetector 346
is linked to a central processing unit (CPU)(not shown)
for processing the sensed data in accordance with the
above-described principles of this invention.
Preferably, the beams of light 340A, 340B from the ~.~Ds
340, 342, respectively, are passed through an optical
mask 343 in order to realize the illuminated strips of
the desired dimensions.
In order to capture reflectance samples with high
accuracy, it is important that the photodetector capture
reflectance data uniformly across the illuminated strip.
In other words, when the photodetector 346 is positioned
centrally above the light strip relative to the mid-point
"0" thereof, the output of the photodetector, as a
function of the distance from the central point "0" along
the X axis, should optimally approximate a step function
as illustrated by the curve A in FIG. 17. With the use
of a single light source angularly displaced relative to
the vertical, the variation in photodetector output
typically approximates a Gaussian function, as
illustrated by the curve B in FIG. 17.
In accordance with a preferred embodiment, the two
LEDs 340 and 342 are angularly disposed relative to the
vertical axis by angles ~ and ~, respectively. The
angles ~ and ~ are selected to be such that the resultant
output of the photodetector is as close as possible to
the optimum distribution curve A in FIG. 17. According
to a preferred embodiment, the angles ~ and ~ are each
selected to be 19.9 degrees. The photodetector output
distribution realized by this arrangement is illustrated
by the curve designated as "C" in FIG. 17 which
effectively merges the individual Gaussian distributions
of the light sources to yield a composite distribution
which sufficiently approximates the optimum curve A.
WO91/11778 2050589
43
The manner in which the plurality of light strips of
different dimensions are generated ~y the optical
f c~nhead by means of an optical mask is illustrated in
FIG. 18. As shown therein, the optical mask 350
essentially comprises a generally opaque area 352 on
which two slits 354 and 356 are defined for allowing
light from the light sources to pass through so as to
illuminate light strips of the desired dimensions. More
specifically, slit 354 correcpon~ to the wide strip used
for obt~ining the reflectance samples which correspond to
the characteristic pattern for a test bill. According to
an illustrative emhoAiment~ the wide slit 354 has a
length of about .300" and a width of about .050". The
second slit 356 is adapted to generate a relatively
lS narrow illuminated strip used for detecting the thin
borderline surrounding the printed indicia on currency
bills, as described above in detail. According to the
illustrative embodiment, the narrow slit 356 has a length
of about .300" and a width-of about .010".
It will be obvious that high precision machining
would be required for precisely defining the slits. In
practice, it becomes difficult to machine the narrow
strip 356 on the optical mask 350. This problem is
approached, according to a preferred embodiment, by
defining the mask 350 in the form of separate sections
360 and 362. The section 360 has one edge machined to
correspond to one half section 356A of the desired slit
356. The second section 362 has a corresponding edge
machined to correspond to the other half 356B of the slit
3S6. When the two sections 360 and 362 are mechanically
linked together, they effectively define the narrow strip
356. The advantage with this approach is that, the two
halves 3S6A, 356B, which together define the strip 356,
can be precisely defined since machining on the edges of
the mask can be handled with much more precision than
machining within the mask itself.
