Research On Moving Objects Detection Algorithm In Video Sequences

At present, the study of moving objects detection technology has gradually become a frontier research subject. It plays an important role in many field, such as military navigation and intelligent monitoring. On the basis of a large number of reference literature at home and abroad, this paper mainly analyzes the moving objects detection algorithm for static background.First, this paper introduced the relevant knowledge of digital image processing, laid a theoretical foundation for the analysis of moving objects detection algorithm. Next, analyzes some common algorithms, discusses the advantages and disadvantages of various methods and their limitations. Aiming at the fact that the traditional Gaussian mixture model can not deal with the mutations effectively, an improved method is proposed. The improved method improves the model update mechanism which uses a learning rate and updates the variance using the adapting rate, so that it can adapt to the disturbance of local light. It also updates the mean selectively based on the rule of frames for the global abrupt illuminance change.Due to noise, shadows and other factors, the edge of the target by Gaussian mixture model algorithmsome is not clear enough. To sovle this problem, this paper proposed a new moving object detection method based on edge detection and Gaussian mixture model. Use the improved method which has introduced before to obtain a preliminary moving object. Canny edge detection operator is used to get the edge figure on the space domain, and extract the preliminary outline of the moving object. Using the multi-resolution characteristics of the image-pyramid and further operation, extract another preliminary moving object again. Obtain the result by logic and processing for these two objects. Because the impact of environmental changes on the edges of objects is very little, the new methed which combining with edge detection could extracte the clearer and more complete moving object image. The experiments for different indoor and outdoor video sequences, show the robustness and practicality of the proposed algorithm.