Research And Application Of Moving Target Detection And Shadow Elimination Based On The Background Update

Because endangering public security events occur frequently in recent years, more and more researchers pay attention to intelligent video monitoring as one of the core technology in the field of public security. However, many researchers perform in perfect conditions at present, and do not take into account time complexity of algorithm, so the accuracy of moving target detection is vulnerable influenced by the outside environment, such as illumination change, leaves dithering, lens rock, etc. Many algorithms are very difficult to put into use in real-time video monitoring because of the reason which the time complexity of algorithm is very high. Therefore, a suitable moving target detection algorithm is one of the major problems of the present study.Moving target detection is not only the basic premise of moving target tracking, classification and identification, but also the basic work of sequence image analysis. Therefore, moving targets detection and shadow technology is the most necessary steps in video processing. Now moving target detection mainly experiment in two kinds of scene, one is the scene of background moving, the other is the scene of background fixed. This thesis mainly focuses on moving target detection and shadow elimination in the fixed scene background, the main work includes the following aspects.First of all, this thesis mainly introduces the three kinds of commonly used moving target detection algorithm. It first briefly introduces the principle of three kinds commonly used detection algorithm, which has been conducted in the actual environment, analyzes the advantages and disadvantages of each algorithm by the experiment, and provides some basic theories of next suggested algorithm.Secondly, it puts forward moving object detection algorithm based on the background difference method combined with frame difference method. The method is based on the advantages and disadvantages of background difference and frame difference, uses the median method to establish the background model, updates background model by combining with frame difference and background difference, then compares the differences of current frame and background, uses the morphological method to get moving target size and position. The method compares with frame difference method in order to verify the effectiveness and rationality of the proposed algorithm.Thirdly, it introduces a kind detection method of using neighborhood pixels to build background model. When a pixel is judged as a background pixel, it needs to use the pixel to update background model. But if illumination change or the car stop and leave, it can produce the ghost in the update background. So this thesis puts forward a method to quickly eliminate ghost and does not produce larger hole in moving target that different area select different random probability to update background. Comparing with the frame difference method the detection method is very effective.Fourthly, this thesis introduces some common used shadow detection methods, and the shadow detection method of combining RGB color feature and texture feature are proposed. The video target segmentation is achieved by background difference and multi-frame difference. Then, the RGB similarity (RGB vector angle become smaller) in shadow area and the Local Binary Pattern (LBP) is employed to detect and eliminate the shadow.Fifthly, the moving target Shadow elimination algorithm is applied to the multiple targets monitoring system. We analyze of the function and framework of the system, realize the multi-objective monitoring system, detect algorithm robustness in the complicated environment, and eventually make the system products and extend the intelligent video monitoring neighborhood.