Research On Methods Of Multiple Objects Tracking In Intelligent Visual Surveillance

Intelligent video surveillance system is intended to apply computer vision techniques to video surveillance system in order to process the video data for analyzing, estimating, detecting, tracking and recording the object’s behaviors.Multi-target tracking technique is the essential part of intelligent video surveil-lance system, which includes computer image processing, pattern recognition, ar-tificial intelligence and automatic control. This technique has a promising appli-cation prospect and research value.Multi-target tracking technique in intelligent video surveillance system re-quires a stable operation of the system which means the adaptability to various scenes, high robustness and instantaneity are becoming the hottest issue in this field.The major research approaches and innovations in this paper are:This paper proposed a combination approach of Frame Difference and back-ground subtraction, this paper’s research method benefits from the advantages of both these two approaches, such as the adaptability to illumination, the detection of a complete moving target, then use Frame Difference approach to update the self-adaption backgrounds.This paper proposed a self-adaption approach. This paper has a binaryzation on the moving regions to separate moving targets. Both theoretical analysis and experiment have proved this approach’s effectiveness in improving detection rate and robustness.This paper proposed a multi-feature extraction approach to calculate the tracking centers and the each features’corresponding weight, which will form joint features to complete the target tracking.