The Light Processing In The Moving Target Detection

In recent years, with the rapid development of digital image processing and digital circuit, the intelligent video surveillance has been widely used in all aspects of our life. Compared with the traditional monitoring, the intelligent video monitoring systems has the advantages of saving time and labor. Although there are some achievements in the research and application of intelligent video surveillance systems at present, but there still exists some problems, such as the great influence of light on the moving target detection, which could bring the shadows, the shadows will affect the moving target detection. Background modeling and object detection algorithm is the basic theoretical knowledge of this research. How to build the background model and moving object extraction algorithm is the key to solve the problem. The good model at present are mixed Gauss model and ViBe model. Compared to the mixed Gauss model, ViBe model has the advantages of real-time efficiency and less memory, therefore, this thesis selects the Vi Be model as the theoretical basis of model experiment. Based on the ViBe model, the work which has been done in dealing with the shadow and the ghost are as follows:(1)The thesis has made a research on the pre-processing, morphological processing and image logical operations. In the pre-processing, the thesis introduced the expansion, corrosion, opening operation and closing operation. In the logical operations, the thesis introduced “and”, “or”, “not”.(2)On the ghost problems in the object of ViBe model, the thesis introduced the binarization and edge extraction. Based on the theory of Ostu, we did the binarization on the current frame corresponding to the foreground object, then we did the the logic “and” operation between the foreground object the image, which has been through the binary processing, which can effectively eliminate a part of ghost. In order to handle the remaining ghost, we die the edge extraction of Sobel on the current frame, then we did the expansion on the edge image, at last we did the logic “and” operation between the expansion image and remaining ghost image, which can finally eliminate the ghost.(3)On the shadow problems in the object detection of ViBe model, the thesis introduced the HSV color space and the average method. The images what we need are the RGB images corresponding to the current image and background. But there exists no background, so we try to use the average method to obtain the background. Then we did the HSV transformation on the two RGB images. By using appropriate threshold corresponding to the “H”, “S”, “V”, we can eliminate the shadow.(4)On the problems that the ghost and shadow exist at the same time in the object detection of ViBe model, an improved ViBe algorithm is proposed, the improved ViBe algorithm has a certain inhibitory effect on ghosts and shadows, which can be seen in the numerical experiments.