Research On Motion Target Detection And Shadow Suppression Algorithm Based On Video Sequence

Recently, the moving target detection has been widely used in industrial inspection, traffic navigation and intelligent systems and many other fields. Moving target detection plays a key role for the follow-up study of target recognition, classification, tracking and behavior understanding etc. The shadow of object is detected as a part of moving target mistakenly, which leads in that the accuracy of target detection result decreases, we must do the shadow suppressing for the moving target. Therefore, there is a practical significance to study the algorithm of this paper.Firstly, this paper studies the development situation of moving target detection and shadow suppression algorithm at home and abroad, and analyzes research value and significance of algorithm. The classic algorithms of target detection which includes optical flow method, frame difference method and background subtraction method are studied in the paper. The paper also compare and analysis of the median filtering method, average filtering method and Gauss model for background modeling method.Secondly,this paper focuses on the analysis method of Gaussian mixture model. This paper analyses the shortcomings and deficiencies of the traditional Gaussian mixture background modeling, and proposes an improved GMM method. For the confirmed update rate ?, this paper proposes the concept of scene change rate which can change update rate ? adaptively. For the situation that motion state of moving target is changing, this paper proposes a method which combine GMM method and frame difference method which can judge the movement of the video. For the error detection in large area resulted from sudden illumination changes in the scene, this paper introduce the normalized cross correlation to measure the correlation between the current frame and the background frame, which can judge and suppress light mutation phenomenon. The experimental results show that the algorithm in this paper can be better solve the problem better than the traditional algorithm.Finally, analyses the property of shadow and compare the principle, advantages and disadvantages of current popular shadow suppression algorithm, the paper proposes a shadow suppression algorithm based on Gaussian mixture model. The algorithm uses the property of shadow in HSV color space to judge whether the moving foreground is suspected shadow model, then using Gaussian mixture shadow model(GMSM) to cluster suspected shadow, finally completing shadow suppression. The experimental results show that the algorithm in this paper can more suppress shadow effectively and has a better real time character. The experimental results verify the validity of the algorithm.