Research On Moving Target Detection Algorithm Under Non-stationary Background

Non-stationary background aims at the movement of the background. It will affect the effect of detection and reduce the accuracy of detection. It is one of the major challenges in the moving target detection. Under the non-stationary background, the movement of background, especially violent waving which is random and has a variety of modes for which the distribution of data has range and whose interior also contains noise. It is very difficult to build a good background model. So it is meaningful to study the problem of moving target detection under non-stationary background.In this paper, for the problem of false detection rate of the moving target detection in the non-stationary background is too high, we proposed the moving target detection under non-stationary background algorithm based on the Gaussian mixture model. Our work contents are as follows:(1) Anglicizing the mechanism of the non-stationary background. We analyze the mechanism of non-stationary background by analyzing the relationship of time and space between the non-stationary background and the pixels, and note that the key factor which impacts the accuracy of moving target detection is the small probability events in the non-stationary background.(2) Integrating the raw data. This paper analyzes the necessity of data integration, and points out the advantages of data integration by comparing the experiment.(3) We improve the Gaussian mixture model, making it can tolerate the small probability events in non-stationary background.(4) We introduce a vector parameter to each Gaussian distribution in order to distinguish between the small probability events in background and foreground and reduce the miscarriage of justice of background.The proposed algorithm can handle the moving target detection under non-stationary background, and reduce the false detection rate of the moving target detection in the non-stationary background. The result of comparative experimental show that the algorithm improves the accuracy of the moving target detection under non-stationary background.