Supervised classification of type of crowd motion in video surveillance system

Automated surveillance is of vital importance in public places which has large extent of dynamics to be addressed. The complexity of analysis of such surveillance increases as the size of crowd goes on increasing. This paper attempts to propose an algorithm to analyze and classify the type of motion in a crowd. The analysis is based on texture analysis of video sequence. Nearest neighbor classification is used to classify the motion into predefined classes. The algorithm is tested on standard PETS database.