Review of the Application of Matrix Information Theory in Video Surveillance

Abstract

Video Surveillance is an active research topic in Computer Vision. Its applications include authentication in security sensitive areas, biometric-based specific person identification, overcrowding statistics and congestion control, strange situation detection and alarming, interactive surveillance using multiple cameras and so on. Video surveillance mainly involves modeling of background, detection of motion, classification of moving objects and object tracking. Background modeling is often used to detect moving object in video acquired by a fixed camera. Subspace learning method namely Principal Component Analysis (PCA) have been used to model the background to represent online data content while reducing dimension significantly. Detection and classification of the object of interest in the image captured by the camera is a vital step for automatic activity monitoring. Linear discriminant analysis (LDA) is a well-known classical statistical technique for dimension reduction and feature extraction for classification. This chapter gives an overview of the applications of Matrix Information Theory in video surveillance viz., background modeling by PCA, face recognition/object classification by LDA and finally object tracking using a covariance-based object description. The algorithms which are discussed in this paper are somewhat related to the broad area of Matrix Information Theory.