Research On Moving Object Detection And Shadow Elimination Algorithm Based On GMM

In recent years, the video surveillance system has developed rapidly,and has been widely used in many fields, such as transportation, banking,military security and so on. Moving object detection in complex scenes is a hot research topic in the field of intelligent video surveillance, which lays the foundation for the follow-up research of target tracking and intelligent alarm. In this paper, we focus on the research of the static scene target detection technology, and focus on the mixed Gauss background modeling and moving object shadow elimination. The main work is as follows:Aiming at the problem that the parameters of traditional GMM(mixed Gauss model) are rough, which leads to the large deviation of background model estimation, a K-EM algorithm is proposed. Firstly,the video frame image is obtained by morphological denoising, and then the HSV color space is introduced, and then the HSV mixed Gauss background model is established, and the K-means and EM algorithm are used to initialize the parameters. Experimental results show that the algorithm has higher foreground detection rate under the condition of outdoor illumination, but the effect of shadow suppression is not obvious.In the aspect of shadow elimination, this paper proposes a new algorithm based on color texture and fuzzy integral. Firstly, color clustering is performed on HSV color space, and the rotation invariant LBP homogenization model is used for texture feature. Then, the Choquet fuzzy integral of fuzzy theory is used to fuse the similarity measure of the two, and then compared with the selected threshold. The experimental results show that the algorithm has a better detection effect on the target and the background, and the detection effect is not obvious for the object with small target or similar background color.Aiming at the deficiency of the double feature shadow elimination algorithm, this paper proposes a new algorithm of shadow detection based on PC A (principal component analysis) and Choquet integral multi feature fusion. The main feature of the PCA of the shadow are extracted,and then the Choquet fuzzy integral fusion of multi feature background model, finally extracts the moving target pixel of the image by means of Ostu algorithm and an adaptive. The experimental results show that the results of the background modeling and shadow elimination algorithm in this paper have high accuracy, strong anti-interference ability, and can adapt to different scenes and environments.