Television – Object tracking
Reexamination Certificate
2000-02-22
2004-03-16
Diep, Nhon (Department: 2613)
Television
Object tracking
C348S077000
Reexamination Certificate
active
06707487
ABSTRACT:
FIELD OF INVENTION
The present invention relates to systems for tracking the movement of multiple objects within a predefined area.
DESCRIPTION OF PRIOR ART
As camera, micro-electronic and computer system technology continues to advance at a great pace there has been an increasing supply of machine vision systems intended to replace well defined, repetitive vision/recognition tasks which had previously been performed by humans. Early systems were designed to recognize parts moving along assembly lines to aid the manufacturing process. More recently, many inventions have been put forth to recognize humans and there movement. The variability of humans and their clothing as well as the complexity of the backgrounds within which they endeavor has presented a significant challenge to state-of-the-art technology. Considerable attention has been paid to various techniques for discerning the human shape from its background using edge detection techniques that look to remove stationary, i.e. background, information. There are two major factors that effect the success of these various techniques. The first is image resolution, which drives the amount of information and hence detail that is available to the accompanying computer system in order to differentiate foreground from background. Ideally the greater the resolution the better. However as resolution increasing so does the cost of the camera as well as the accompanying computer system. Even more importantly, as resolution increases time to process increases significantly. And as time to process increases the ability for these systems to perform real-time operations is impeded.
The following are seven examples of machine vision systems designed in some way or another to recognize human or object motion from within a predefined area.
In November of 1994 U.S. Pat. No. 5,363,297 entitled Automated Camera-Based Fracking System for Sports Contests issued to Larson et al. This system employed multiple cameras to continuously monitor the area of an ongoing sporting event. Each camera would feed its information to an accompanying computer system for analysis that consisted of extracting the players from the stationary background and thereby tracking their silhouettes. The inventor had anticipated problems when the individual players would collide or otherwise engage each other and hence merging their individual silhouettes. A need was also recognized to initialize the system by first identifying each player as they appeared within the system's field of view. Larson et al. specified two solutions for these problems. First, they proposed attaching monitors to the tracking system that would be operated by humans who would perform the initial recognition as well as all subsequent re-identifications when the system lost track of a player due to a merging of silhouettes. Secondly, they proposed attaching electronic tracking devices and employing triangulation of received signals in order to identify and track individual players. There are at least four major problems with the Larson patent. First, the amount of digital processing required to perform the player extraction in real time greatly exceeds the cost-effective computer technology of today, let alone 1994. Secondly, in order to perform this extraction, a greater amount of detail would be necessary which would thereby increase the cost of implementation by requiring more cameras and related computer systems. And of course the additional detail would only tend to further slow down the responsiveness of the system. Thirdly, the requirement of one or more operators to initially recognize and then re-identify players is extremely limiting and costly. This requirement essentially made the patent economically impractical for monitoring non-professional youth sporting events, where the system cost including the ongoing cost of the human operator would greatly exceed smaller revenue streams. It should be noted that this operator would more than likely be a parent of one of the youths who would probably be unfamiliar with all of the players and who would more than likely find it stressful to make so many decisions in real time. This approach would also require training and retraining operators, which would also be prohibitive. Fourthly, the type of electronics necessary to track players in real time would have to operate at higher frequencies that would also mean that it would be more expensive, providing a further economic drawback. The fifth major problem is that the system could have a difficult time determining the orientation of the player. For example, while Larson's invention could detect the direction of an individual's motion, it could not determine if the person was facing forwards or backwards, let alone if the person's head was turning in another direction.
In December of 1995, U.S. Pat. No. 5,473,369 entitled Object Tracking Apparatus issued to Keiko Abe. This system was concerned with the actual image processing techniques used to follow an object from frame to frame. The inventor described prior art that compared the images block by block from one frame to the next, where a block is assumed to be one or more pixels of the image. It was pointed out that such systems depended upon time consuming and error prone statistical calculations that were especially susceptible to misinterpretation when the object changed size within the field of view or disappeared altogether. Abe proposed taking the same video frames but first separating them into luminance and color histograms that are then to be compared frame by frame. By comparing the histograms rather than blocks Abe argued that the system would be more accurate and efficient that the block matching systems. However, there are at least five major problems with Abe's patent. First, the effectiveness and reliability of this technique is highly dependent upon the lighting conditions initially and over time within the field of view being tracked. For instance, if the initial frame was taken under a well-lit condition, the luminance histograms of the object may be ideal. However, when the lighting conditions are poor to begin with, or worse yet change from frame to frame as might happen with sudden burst of ambient light, the luminance histograms will be subject to considerable error. Secondly, relying upon color histograms is equally uncertain due partly to the susceptibility of color detection to lighting conditions, which again may vary from frame to frame, and in part to the potential for object and background blurring when color schemes overlap. A third problem is that Abe's system does not lend itself to tracking multiple objects that may have similar or identical luminance/color information and which may overlap from frame to frame. A forth problem is discussed in Abe's specification that indicates a requirement of a human operator to initialize the system by selecting the portion of the video image which contains the object to be tracked, his so-called region designating frame. This requirement would be even more restrictive when consideration is given to multiple object tracking where the objects may go in and out of the field of view of temporarily overlap each other. And finally, a fifth problem is alluded to in the specification where it is expressed as an opportunity of the system to automatically control the pan, tilt and zoom of a camera. In so doing, Abe states that the system is capable of “coping with any change in the in the size of the object and which can photograph the target object always in a desirable size, thereby attaining a substantial improvement in terms of the facility with which the apparatus can be used.” Hence, it is recognized that this method/apparatus is still very resolution dependent similar to the block methods it is attempting to improve upon.
In April of 1997, U.S. Pat. No. 5,617,335 entitled System for and Method of Recognizing and Tracking Target Mark issued to Hashima et al. This invention is attempting to address the problem of determining the three-dimensional co
Aman James A.
Haller William R.
Diep Nhon
In The Play, Inc.
LandOfFree
Method for representing real-time motion does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for representing real-time motion, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for representing real-time motion will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3242424