Smart video-based surveillance: Opportunities and challenges from image processing perspectives

Traditionally CCTV security and surveillance systems are meant for deterrent rather than for prevention where these systems will simply record scenes as viewed by video cameras regardless whether there is any activity or not. In the event of some criminal prosecution procedures, these recorded videos may be served as evidence in the court of law. The 9/11 incident has changed the way on how these video-based surveillance systems are being employed. More automated functions and “smart” algorithms have been incorporated into such systems so that these systems can detect moving object, identify to some degree what the object is and possibly tracking its movement. However, the problems of object identification, recognition, and tracking are far from being solved. The London bombing and the recently Boston bombing are proof that automatic system failed to identify the suspect. In fact, it took hundreds of man hours for the authority to finally identify the criminals. In this talk we will look at emerging smart vide-based surveillance systems. The talk will not focus on any particular model, or system or algorithms. Instead, the talk will concentrate on what are the opportunities and challenges faced by the existing systems. In particular, we will be exploring these from the perspective of image processing. Thus the talk will serve two purposes. Firstly, it will give a brief overview on what smart video-based surveillance is and secondly, it will present the focus areas for research activities by highlighting the challenges. To be specific, these areas of interest that will be covered by this talk will be on Face Recognition, Activity Recognition, Plate Recognition, Object Abandon/Stolen Recognition, and Crowd Detection. It is hoped that this talk will motivate and inspire many other young researches particularly from the ASEAN region to take more serious effort in this exciting field.