Research And Implementation Of Human Movement Detection Method Based On Merging Depth Video And Tracking-Learning

Motion detection and tracking of human body is the key technology in applications such as intelligent monitoring and intelligent transportation and has a very important research value. The complexity of the application scenarios and variability on human make high demands to human motion detection and tracking technology. Therefore, building a high robustness, high accuracy, good real-time, human motion detection and tracking system that can solve occlusion and distortion problems becomes a serious problem. In this paper, we research human motion detection and tracking method based on the depth information.In terms of human detection, this paper introduces a method by combining frame difference and background subtraction and the method can solve problems of frame difference and background subtraction, such as failing to detect stationary pedestrians. And, we used this method to the depth video and eliminated the influence of light on the result. Experiments show that the method can enhance the robustness of the system. In this paper, we study and implement TLD algorithm and propose a method to optimize TLD by using depth information. Under the premise of guaranteed real-time algorithms and robust, we optimize the feature point selection strategies and track frame scale change strategy though by modifying and improving the TLD algorithm based on depth information. Experimental results show that the Improved TLD has a better performance on human detection and tracking. In addition, we established a depth video database and callout static parameters of the database. In the processing of establishing the database, we studied to remove the image noise in depth image and achieved good results.Finally, we designed and implemented a real-time human detection and tracking system based on depth video, which achieves detection and tracking methods that we have studied.