Construction of intelligent intrusion detection system based on KINECT

Surveillance cameras mainly capture the video footage of a person. It is not capable of identifying suspicious intruding people when it is dark. Person identification using gait features is an area investigated for long. But extracting gait features from video data is computationally complex, and real-time identification is difficult. In this work, we use KINECT developed by Microsoft to track people's mobility features and obtain the characteristic features to identify (classify) the person. KINECT can deliver many features useful to identify a person, including simple static features like height, hand-span, etc. to complex dynamic features like hand movement, leg movement, etc. In this work, we used such features for person identification and thereby intrusion detection. Actual data was collected from subjects and classified using Artificial Neural Network (ANN). A very low False-Acceptance-Rate (FAR) and False-Rejection-Rate (FRR) could be achieved.