Research On Object Detection And Tracking In Intelligent Surveillance System

With its significant practical value and broad application prospects, intelligent surveillance system is an important issue within computer vision field, it is a challenging task which involves the automatic and intelligent detection, tracking, classification and behavior recognition of video sequence. In this dissertation, we focused on two aspects of intelligent surveillance system: object detection and object tracking, discussed the related problems of object detection, feature extraction, object tracking, occlusion handling, etc.In terms of object detection, we improved frame differential method, using the differential result to update motion history image to detect moving objects.In terms of feature extraction, we analyzed geometrical and appearance features, introducing color histogram feature and Bhattacharyya distance as its measurement.In terms of object tracking, we investigated two tracking algorithms: Kalman filter based multi-object tracking and Particle filter based multi-object tracking, discussed the principles and algorithmic steps, improved the two algorithms and presented the experiment results, respectively.Finally, we realized a video surveillance system using VC++ integrates the methods discussed in this dissertation to verify their validity.