Research On Moving Object Detection And Tracking In Video Sequence

In recent years, Moving object detection and tracking in video has become an important issues in computer vision which is getting more and more attention. It first detected a moving target from the surveillance video, and then it according to the corresponding algorithm does some real-time tracking. However due to the complexity and uncertainty environment and the target’s own diversity limit the performance of existing algorithms. Therefore design a robust, real-time, reliable and stable moving target detection and tracking algorithm is still a very challenging task. The main works of this paper is to study and analysis the main algorithm of the moving target detection and tracking, and proposes a new moving target detection method based on three-frame difference method and threshold segmentation and improved Camshift tracking algorithm based on multi-feature fusion.Firstly focus on the problem of traditional moving target detection algorithm (Inter-frame difference method, Background subtraction method and optical flow method), based on which proposes a new moving target detection method based on three-frame difference method and threshold segmentation. First the one dimensional cross entropy threshold segmentation and two-dimensional cross entropy threshold segmentation method are discussed in greater detail and proposes the fast two-dimensional cross entropy threshold segmentation algorithm, second adopting three-frame difference method to get the moving target region, then use the fast two-dimensional cross entropy threshold segmentation algorithm to segment and use some morphological operations to get the moving object extraction. This method can satisfy the real-time, while accurately and efficiently detect moving targets.Secondly in view of the traditional continuous adaptive mean-shift algorithm (Camshift) mainly use the color information of the object to track the object, which may easy to cause the target tracking problems such as tracking lost or misplaced under the condition of drastic illumination changes, severe target occlusion. To solve this problem, Proposes a improved Camshift tracking algorithm based on multi-feature fusion. First a new two-dimensional probability distribution histogram is proposed based on the hue component H and the saturation component S in HSV color space and texture feature based on gray level co-occurrence matrix are used to build target model. Second calculate the degree of color similarity and texture similarity between current frame the candidate model and the target model by Bhattacharyya coefficient and Mahalanobis distance. Then a new convergence criterion is proposed based on Bhattacharyya coefficient and Mahalanobis distance. The algorithm effectively solves the problem of tracking object lost or misplaced under illumination change or target occlusion.