Detection And Tracking Of Moving Objects With Complicated Conditions In Video

Nowadays, there are wide applies and great actions of video systems in national economy and national defense. The research hot points are how to improve intelligent ability, extend applied area and advance system performance. Video processing with complicated conditions is a prominent aspect, and its results will affect high-level understand of video. The thesis is focused on motion detection and tracking with complicated conditions which include illumination change, shadow and occlusion.Firstly, the pre-processing methods that include removing both noise and illumination change are studied. Based on review on conventional pre-processing methods, two new methods are developed. One is an adaptive vector mixture filtering method, which can remove gaussian and impulse noises adaptively. The other is illumination change detecting, which can remove the effect of illumination change by putting the gene factor of illumination change into the model of background.Secondly, the methods of motion detection with shadow are studied. A modified mixture ganssian model of background subtraction model is developed, which is updated by the gene of illumination change, and results in removing the effect of illumination change. Also, an optimal selection method is addressed, which is for finding both the best color space and a shadow eliminating operator in order to remove shadow maximallyThirdly, active contour-based tracking methods are studied, which include contour detecting, contour tracking and occlusion removing. With contour detecting, the improved Chan-Vese model is presented to obtain the initial contour of tracking. With contour tracking, the improved fast marching method in Level Set theory is provided in order to tracking with illumination change. With occlusion removing, the occlusion detecting method with geometry and histogram information is addressed to eliminate the effect of occlusion.Finally, a motion detection and tracking experimental system with complicated background conditions is designed, which provides an experimental fiat to validate relative algorithms. It specifies the main frame of system, the functions of every model, estimating methods and limitations. The experimental results indicates that the system is certainly valuable in stated range.