Moving Target Detection Technology In The Complex Environment

Moving object detection is an important researching area in computer vision community, which has widely application in industrial, medical, military, education, business, and sports areas. Exact detection of moving object is the basic work of image sequence analysis, but also the hotspot that is studied by many contemporary scholars at home and abroad.In this paper we do an in-depth study about moving object detection technology and its implementation in a complex environment. The main contributions are as follows:Firstly, we analyze the process of moving object detection based on difference image, and give the specific implementation of object's detection and extraction. Also we give a double threshold method of difference image segmentation.Secondly, followed by experiments on existing background model, such as mean background model, subsection background model, median background model, Gaussian background model and random background model, we have made systemically analyzing about them. Beside, we also summarize the advantage and disadvantage of them. In view of the advantages of various background models, we present the integration strategy of multi-background, and give the realization of the strategy.Finally, by investigating image's watershed segmentation algorithm, we propose a method of detecting moving object based on watershed. This method divides the whole image into small regions and gets the moving regions through analyzing of small areas. By comparing with single pixel's difference method, it exhibits robustness. We give two forms of its specific realization: one is based on background segmentation, another based on image sequence segmentation. The former divides the background image into different regions and uses some models to describe its changing state. It has rapid processing speed. The latter has to segment every frame, and then extracts the whole object by appropriate region merging steps. Through our testing, this method shows good performance in complex environment.