Research On Motion Target Detection Methods Based On Time And Space Significance

Disturbed by dynamic background, illumination changes, object is occluded and camera shake, including other external factors, the moving object accurately detected from the video sequence has been a problem in computer vision and image processing. Final image analysis and discriminate results the pros and cons depend on the detection results. From the point of view of computer vision that detected moving object from video sequence, it's also the focus of human eyes will pay more attention. Thus, we proposed a method to detect moving objects that combining with the bottom-up and top-down computer vision models, this method effectively utilized the superiority of spatial saliency detection algorithm can simplify complex scenes, and applied it to the moving object detection algorithm. Steps of our algorithm are that:first detected spatial saliency in video sequence, then add the time dimension, integrated the optimized method of three difference algorithm, obtained the moving objects which spatial-temporal properties were significantly. This method detect salient object from static images converting to the video sequence for salient moving object detection, meanwhile, it can detect moving objects within a short period of time.In this paper mainly work includes:(1) Designed a framework for salient moving object detection algorithm, and analyzed the theoretical knowledge related to our algorithm, then intuitive understanding on knowledge to simulate and experiment them.(2) Through the analysis the human visual system of information feedback mechanism and the bottom-up and top-down models are based on visual attention, explored the spatial-temporal saliency via a variety of algorithms model relating to visual attention mechanism, an algorithm has been designed to detect moving objects which spatiotemporal characteristics are significantly.(3) Selected groups of public video data as test samples, which have many challenges to detect exact object in the international and those scenes are complicated. Compared our method with the current widely used for moving object detection algorithm. Experimental results utilized quantitative and qualitative evaluation, to explore experimental results whether consistent with the expected desired.In this paper, our proposed method which combine with moving object detection and the spatial saliency detection to obtain salient moving object, the experimental results shown that our method can effectively overcome the influence from complicated dynamic background interference, and the results also indicate the method has better robustness when illumination conditions changes and camera is shaking, and only to detect moving objects which spatiotemporal characteristics are significantly.