Image Analysis for Video Surveillance Based Traffic Monitoring System

Visual analysis is concerned with detection and analysis of a target object in a sequential stream of images. While many algorithms are successful to detect objects effectively in controlled environments, they get bugged with the variations of the object’s appearance or with the surrounding change. It happens as the existing algorithms employ fixed appearance surroundings for the objects. Such systems are trained using appearance of the data available before the surveillance begins, which limits the range of modeling, and ignores a large area of information (eg: shape change or lighting conditions) which becomes dynamically effective during surveillance. Traffic-video- surveillance system that incrementally learns a subspace representation, with accordance to the changes in the appearance of the target. The proposed system is decomposed into two independent sub-problems. The first problem is to detect foreground objects on a frame-by-frame basis. It is done by labeling each concerned pixel in the image frame as object foreground or object background. The second is the coupling of object observations at different points in an inter-related sequence to extract the moving objectʹs trajectory.