Image analysis – Applications – Target tracking or detecting
Reexamination Certificate
2000-06-13
2004-02-03
Boudreau, Leo (Department: 2621)
Image analysis
Applications
Target tracking or detecting
C382S209000
Reexamination Certificate
active
06687386
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a monitor apparatus using an imaging unit, or in particular to an object tracking method for automatically detecting an object intruding into an imaging field or image pickup field from a video signal inputted from the imaging unit and automatically tracking the motion of the detected object and an object tracking apparatus for adjusting the imaging direction in accordance with the detected motion of the object.
A video monitor apparatus using an imaging unit or image pickup unit such as a camera has been widely used. A monitor system using a video monitor apparatus is in demand, in which an intruding object is automatically detected from the image input from an image input means such as a camera, the motion of the object is automatically tracked and a predetermined announcement or alarm action can be taken, apart from the manned monitoring system in which an intruding object such as a man or an automotive vehicle entering the monitor field is detected or tracked by a human monitor while watching the image displayed on the monitor.
For realizing such an unmanned system, the first step is to detect an intruding object in the view field by a so-called subtraction method or the like. In the subtraction method, the input image obtained by an imaging device such as a television camera (hereinafter referred to as the TV camera) is compared with a reference background image prepared in advance, i.e. an image not including the object to be detected to determine the brightness (or intensity) difference for each pixel and detect an area with a large difference value as an object. The part of the input image (hereinafter referred to as partial image) corresponding to the position of the intruding object detected in this way is registered as a template, so that a position associated with the maximum degree of coincidence with the template image is detected in the sequentially input images. This method is widely known as the template matching, and is described in detail, for example, in the book entitled “Introduction to Computer Image Processing” edited by Hideyuki Tamura, published by Soken Publishing Co. pp. 149-153, 1985. Further, this method is disclosed in “Digital Picture Processing” published by ACADEMIC PRESS pp. 296-303, 1976 and U.S. Pat. No. 5,554,983, the disclosures of each is hereby incorporated herein by reference.
The aforementioned method of tracking an intruding object by template matching poses the problem that with the change of the orientation or posture of the target object (when the target object person turns to the right or turns around, for example), the deviation of the target object from the matching position increases to such an extent that the accurate and stable tracking becomes impossible.
Specifically, the template matching has the property that the pattern portion high in contrast in the template images coincides for matching. In the case where a vehicle is a target object, for example, substantially the whole vehicle facing forward first constituting an object of matching (the input image
802
in
FIG. 8
) may subsequently proceed in a different direction (orientation) and turn sideways, in which case only the front portion of the turned vehicle constitutes the object of matching. As compared with when the whole vehicle has been the object of matching, the center of matching changes to the front portion of the vehicle and therefore the displacement occurs such that the center of the detected location is deviated from the center of the vehicle to the front portion thereof.
This will be explained with reference to FIG.
8
.
FIG. 8
is a diagram in which a vehicle running along a curved lane within an imaging field is assumed to be an intruding object for explaining the flow of the tracking processing. Numerals
801
a
,
803
a
,
8
OSa,
807
a
designate template images at time points t
1
−1, t
1
, t
1
+1, t
1
+2, respectively; numerals
801
,
803
,
805
,
807
designate images indicating the updated template images
801
a
,
803
a
,
805
a
,
807
a
, respectively; numerals
802
,
804
,
806
,
808
designate input images at time points t
1
, t
1
+1, t
1
+2, t
1
+3, respectively; numerals
802
a
,
804
a
,
806
a
,
808
a
designate the positions of the object detected by template matching at time points t
1
, t
1
+1, t
1
+2, t
1
+3, respectively; and numerals
802
b
,
804
b
,
806
b
,
808
b
designate the positions of the template images in the immediately preceding frame (i.e. the template images at t
1
−1, t
1
, t
1
+1, t
1
+2, respectively).
In
FIG. 8
, the template image
801
a
registered at time point t
1
−1 is an image of a moving vehicle with the front portion thereof directed substantially forward. At time point t
1
, the template matching is conducted using this template image
801
a
to detect the position moved by the target object, while at the same time updating the template image
801
a
to the template image
803
a
. Then, at time point t
1
+1, the template image
803
a
is updated to the temperate image
805
a
. Further, at time t
1
+2, the template image
805
a
is updated to the template image
807
a
. By conducting this process up to time t
1
+3, the template matching which has thus been conducted on the front portion including the vehicles' lights at the tracking starting time t
1
is deviated to the left of the vehicle.
This phenomenon is caused by the fact that the matching is conducted in such a manner as to reduce the deviation of the positions between the input image target for template matching and the image portion high in contrast in the template image. In this case, such a portion is the light of the vehicle. As a result, in the case where the target object turns to the left in
FIG. 8
, for example, the deviation is leftward, and vice versa.
Further, at time t
1
, only the vehicle image is included in the template image
801
a
. With the direction change of the target object and the resulting deviation of the template position, however, the image of the background portion other than the object image occupies into the template image
807
a
. In the case where the tracking is continued using a template image like the template image
807
a
including many images other than that of the target object, the target object cannot be matched and the background portion that has occupied into the template is matched. Thus, in the case where the target object changes in direction, for example, the object tracking method using the template matching cannot guarantee the tracking of the target object and cannot assure stable tracking because the pattern of the target object apparently moves, which causes the position of the template to deviate.
Further, the aforementioned method of tracking an intruding object by the template matching, which is the process of detecting the portion of the input image associated with the maximum degree of coincidence with the template image, poses another problem that the target object, if temporarily hidden behind some blocking object, cannot be found. Still another problem is that in the case where a plurality of objects (moving objects) exist in the view field and the target object is temporarily hidden behind another moving object, the template image may be updated undesirably with the blocking object in front of the target object as a template image, thereby making it impossible to track the target object.
An example in which the target object is cannot be tracked is explained with reference to FIG.
14
.
FIG. 14
is another diagram for explaining the flow of the process for tracking an intruding object in the case where a plurality of moving objects are imaged in superposed relation. Numeral
1801
a
designates a template image of an object updated at time point t
0
−1, numeral
1801
a view showing the position of the template image
1801
a
in the input image at time point t
0
−1, numeral
1802
the input image at time point t
0
, nume
Ito Wataru
Ueda Hirotada
Antonelli Terry Stout & Kraus LLP
Boudreau Leo
Hitachi Denshi Kabushiki Kaisha
Miller Ryan J.
LandOfFree
Object tracking method and object tracking apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Object tracking method and object tracking apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Object tracking method and object tracking apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3326235