Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
- CA 02217366 1997-09-30
FACIAL RECOGNITION SYSTEM
Field of Invention
The present invention relates to facial recognition
systems.
Background of the Invention
Known methods include: asking the target person to stand
in front of a controlled background (e.g. opaque barrier),
while the camera is positioned and adjusted (zoom and focus)
under manual control; use of slow and elaborate software logic
performing post analysis on a captured video frame to isolate,
crop, and enhance the facial representation of the target
person and filter out background artifacts.
Most facial recognition systems available today can
function accurately only when one face is present in the image
submitted for analysis. The reality, however, is that multiple
faces generally appear in the field of view when areas are
being monitored where line-ups or queues occur. Examples of
such situations are line-ups at immigration counters, border
crossing and automated wall machines (cash dispenser, frequent
traveller border crossing booth).
CA 02217366 1997-09-30
Summary of the Invention
The present invention will automatically isolate and track
the face of a person in a defined target area, generate a
digital image of the isolated person's face, remove background
artifacts contained in the digital image and confirm the
person's identity through use of facial recognition logic.
Steps involve: 1) locating the person closest to the camera
in the defined target area; 2) controlling an automated camera
to crop the overall size of the target face to ensure that the
crown-to-chin ratio is within predefined limits; and focus the
camera to ensure the level of detail is within predefined
limits; 3) analyzing the video output from the camera to
automatically detect and remove artifacts from the background
of the generated image, leaving only the facial representation
of the target person in the digitized data frame; and
presenting the isolated facial image of the target person to
a secondary processing step to allow identity confirmation
based on use of automated facial recognition logic.
The advantages associated with the present invention
include: automated capture and tracking of a person's facial
image; automated removal of artifacts from the captured image;
accelerated capture of facial images (based on predefined
optimum and consistent specification; improved accuracy of
secondary processing steps (such as automated facial
recognition logic); reduced operating costs (such as labour
CA 02217366 1997-09-30
costs for operators carrying out manual capture of facial
images); and increased security.
According to the present invention, there is provided a
method for isolating and tracking the face of a person in a
defined target area, generating a digital image of isolated
person's face, removing background artifacts contained in the
digital image and confirming the person's identity through use
of facial recognition logic, comprising the steps of:
a) locating the person closest to the camera in the defined
target area; b) controlling an automated camera to crop the
overall size of the target face to ensure that the
crown-to-chin ratio is within predefined limits; and focus the
camera to ensure the level of detail is within predefined
limits; c) analyzing the video output from the camera to
automatically detect and remove artifacts from the background
of the generated image, leaving only the facial representation
of the target person in the digitized data frame; and
d) presenting the isolated facial image of the target person
to a secondary processing step to allow identity confirmation
based on use of automated facial recognition logic.
Other advantages, objects and features of the present
invention will be readily apparent to those skilled in the art
from a review of the following detailed description of the
preferred embodiments in conjunction with the accompanying
drawing and claims.
CA 02217366 1997-09-30
Brief Description of the Drawings
The embodiments of the invention will now be described
with reference to the accompanying drawing, Figure 1, which
illustrates an embodiment of the invention.
Detailed Description of the Invention
The present invention will automatically isolate and track
the face of a person in a defined target area, generate a
digital image of the isolated person's face, remove background
artifacts contained in the digital image and confirm the
person's identity through use of facial recognition logic.
Steps involve: 1) locating the person closest to the camera
in the defined target area; 2) controlling an automated camera
to crop the overall size of the target face to ensure that the
crown-to-chin ratio is within predefined limits; and focus the
camera to ensure the level of detail is within predefined
limits; 3) analyzing the video output from the camera to
automatically detect and remove artifacts from the background
of the generated image, leaving only the facial representation
of the target person in the digitized data frame; and
presenting the isolated facial image of the target person to
a secondary processing step to allow identity confirmation
based on use of automated facial recognition logic.
CA 02217366 1997-09-30
The invention allows a wide range of processes and systems
which use a person's facial image as part of their business
process, to operate more efficiently and effectively. Steps
1 through 5 above reflect an identity confirmation application,
while steps 1 through 3 reflect use in a variety of wide spread
applications where automated delivery of an isolated facial
image minus background artifacts would provide significant
advantages. Examples of the latter are: capturing a facial
image to include on an identification card (e.g. drivers
license, health benefits card); and capturing a facial image
to determine the acceptability of a person to access a
restricted area.
The invention uses a variety of sensors 4 to detect and
provide position and control information for a camera 3 with
electronic exposure control, motor driven zoom lens and auto-
iris. Control logic 5 uses information from the sensors 4 to
accurately position and focus the camera 3 to ensure that the
closest person 1 is and remains the target within the overall
field of view 2.
Once the camera 3 has been positioned and adjusted, guided
by pre-defined limits, thereby ensuring the target face is the
required percentage of the full video frame and consistent with
the required crown-to-chin ratio, the resultant video frame 7
is processed by analysis logic 8. The analysis logic 8
automatically determines the presence of unwanted artifacts in
CA 02217366 1997-09-30
the image and removes these; i.e. making the background in the
digitized image conform to a predefined pattern and/or
colour/hue.
An altered image 9 minus background artifacts (e.g.
representations of other persons standing near the isolated
person) is then delivered to additional processing step 10.
Examples of additional steps are display of the isolated image
on a video display unit, facial recognition logic or system,
and document issuance system requiring a cropped image of the
person for printing on the issued document.
The present invention enables: automatic location and
tracking of a person's face (within a larger and cluttered
field of view); automated cropping of the person's face
consistent with predefined specifications; automatic isolation
of the primary facial image and the background details;
automatic removal of artifacts in the background that will
negatively effect the secondary process 10 being performed;
automatic conversion of the background in the video frame to
a predefined representation; faster capture of facial images
(for example, avoiding manual adjustments of a camera);
accurate identity confirmation in a crowded environment; and
improved performance of secondary processes such as printing
of cropped facial images on identity documents and secure
access control systems.
CA 02217366 1997-09-30
The present invention will detect the nearest face in a
wide field of view and zoom in on it until this face fills the
screen consistent with pre-defined specifications.
Most facial recognition systems available today can
function accurately only when one face is present in the image
submitted for analysis. The reality, however, is that multiple
faces generally appear in the field of view when areas are
being monitored where line-ups or queues occur. Examples of
such situations are line-ups at immigration counters, border
crossing and automated wall machines (cash dispenser, frequent
traveller border crossing booth). By isolating and focusing
on the nearest person on a target field of view, followed by
removal of background details from the capture video image, a
high quality image may be obtained of a person's face. Any
facial recognition system will benefit from high-resolution
images without interfering objects and consequently the
recognition rate is expected to increase significantly.
The camera system will consist of a high resolution CCD
with electronic exposure control and auto-iris. An ultra-sonic
distance measuring device and an infrared detection sensor will
augment this camera.
Both the ultra-sonic and infra-red detector will be set
up in such a manner that they cover the camera default position
which is the wide field of view (zoomed out). Upon detection
CA 02217366 1997-09-30
of body heat radiation (infrared sensor) and the presence of
it within a pre-determined range (ultra-sonic device) the
camera will be pointed in that direction by means of a pan and
tilt mechanism. Next the camera will be focused and zoomed
based on the distance information obtained from the ultra-sonic
range finder so that the person's face will nominally fill the
field of view. Pre-defined specifications will control the
operation to achieve the required size of a face with respect
to the overall field of view, the required crown-to-chin ratio,
and accuracy of focus.
As the person moves within the tracking range the camera
will track the person and keep their face in focus and zoomed-
in to the required degree.
The camera will have a high resolution CCD sensor with
electronic exposure control and will be equipped with a zoom
lens. This zoom lens has a facility to set it to a particular
magnification setting based on distance information provided
by the ultra-sonic range finding sensor.
Using mirrors in front of the lens driven by small motors
will effect pan and tilt. The lightweight of these mirrors
enables a fast response time and minimizes wear and tear on the
rest of the camera. The motors will be DC motors operated in
a closed servo loop based on signals from the infrared sensor
and indirectly also from the distant sensor.
CA 02217366 1997-09-30
Although at least one patent describes the use of
ultra-sonic and infrared sensors to activate a camera system
the essential difference here is the type of infrared sensor
employed. Unlike the common pyro-electric infrared sensor used
in motion detectors, an altogether different type, thin film
thermopile, sensor is employed here. The essential difference
is that the first kind only responds to movement of an object
that has a different temperature than its surroundings. The
object if cooler, warmer or much hotter than the ambient
environment will trigger this kind of sensor while it moves.
If the object is positioned more or less stationary it will
fail to detect such an object. The second kind differs in the
sense that it can be "tuned" to detect only infrared radiation
emitted by objects having a temperature of a human body (37.5
degrees Celsius). Moreover, and perhaps more importantly, this
type of sensor does not sense temperature variations caused by
movement of an object having a different temperature from its
surroundings but the actual temperature even while the object
is at complete rest. For our intended application, where
people may be more or less stationary while queuing, a sensor
of the first type could not be relied upon to provide the
necessary information whereas a sensor of the second type will.
As mentioned above, a sensor of the second type also will not
respond to object other than those having a human body like
temperature further decreasing the change of "false
triggering".
CA 02217366 1997-09-30
Important elements of the invention are: 1) the types
of sensors and their integration to allow detection and provide
position/control/tracking information for an automated camera
required to isolate and track the face of a target person; 2)
the method for accomplishing pan and tilt; and 3) processing
logic designed to locate, remove and/or replace artifact data
contained in output video from a camera.
Numerous modifications, variations and adaptations may be
made to the particular embodiments of the invention described
above without departing from the scope of the invention, which
is defined in the claims.
