A computerized chess player for classifying movements in security cameras

Prof. Natan Netanyahu from the Department of Computer Science describes his research group's part in developing algorithms for information retrieval and classification of object movements

Prof. Natan Netanyahu from the Department of Computer Science describes the part of his research group in the development of algorithms for retrieving information and classifying the movements of objects such as vehicles, people and animals for automatic analysis of security camera data as part of the "VULCAN" consortium designed to develop video technologies for security and surveillance applications
One of the questions encountered by scientists engaged in basic research is "Why is this necessary?" Prof. Natan Netanyahu from the Computer Science Department at Bar-Ilan University and his research student, Dr. Eli David, also grappled with this question, following their success in developing advanced computer chess programs that learn in an evolutionary way by a genetic algorithm ("Evolution in black and white", Galileo, issue 119, pp. 54-48, 2008)
. As part of this research, Dr. David also participated, among other things, in the World Computer Chess Championship in 2008, in which his Falcon program won second place in the blitz competition.
In a 2004 interview, (see) At the conclusion of the World Computer Chess Championship held at the Brain Research Center at Bar-Ilan University, Prof. Netanyahu observed that the experience gained in the field of computer mind games, and in particular computer chess, may be applied in other fields of artificial intelligence and thereby advance the research and technological front . This vision has indeed come true in recent years, among other things within the framework of the activities of researchers in the Computer Science Department at Bar-Ilan University, who apply techniques developed for computer chess to study and classify the movements of objects such as vehicles, people, animals, etc.
This development was carried out as part of the VULCAN consortium of the Magnat project (generic technological research) which is shared by academia and industry and in which well-known companies such as Elbit, Raphael, Verint, Nice and more participate alongside a number of academic bodies, including the Department of Computer Science at Bar-Ilan University.
Maged VULCAN, which is at the beginning of its fourth year of operation, develops technological building blocks for the analysis and understanding of video for security and surveillance applications. The world of security is networked with video cameras of various types whose purpose is to assist in monitoring and preventing events in real time, and investigating events after they have occurred. For a typical query, such as "find all the people who passed through a certain area of ​​the observed complex in a given period of time", the system should return the relevant data as accurately and efficiently as possible. The main technological problem faced by the head of the association is the excess of visual information currently received from a large number of cameras placed at various sites such as airports, sensitive facilities and even in entire cities (such as London) and insufficient ability to analyze it by human factors. To this end, the group develops technologies to release the "bottleneck" so that the human factor can focus on making operative decisions based on the processed information. The ability to exhaust the information produced in a relevant time is of course conditional on the development of sophisticated algorithmic capabilities and their efficient implementation. The developed technologies should answer a variety of problems in the form of developing various attractive products for the security market.
According to Prof. Netanyahu, the association's technological goals are very extensive; To achieve them, it is advisable to map the "big problem" into a large number of small sub-problems and focus on them first. For example, Bar-Ilan researchers work in collaboration with the Jerusalem start-up company ProTrack on the problem of classifying movements while studying characteristic groups (extracted from video data) by a genetic algorithm. The classification rate achieved so far exceeds 90%.
In addition to the movement classification problem, Prof. Netanyahu and his Ph.D. research student, Lior Friedman, work within the association in collaboration with the RTC company on classical algorithmic problems, such as "finding the nearest neighbor" in multidimensional Euclidean space and general metric space, in the context of VULCAN data. This well-defined problem is a mapping of the original problem that deals with complex queries that one wishes to present to the computerized system. By solving the mapped problem, it will in principle be possible to get relevant answers from the original video data database, za, algorithms of this type will help in retrieving information from a huge amount of video data from the security and monitoring cameras.
Maged VULCAN engages in challenging research in the fields of computer vision, such as tracking objects between cameras and in a crowded scene, classifying objects into subclasses, identifying behavior and abnormal behavior, with an emphasis on the field of security and according to the needs of the Maged.