Moving Target Detection And Tracking Based On High Speed Vision

High-speed motion exists in the field of national defense and military.It has the characteristics of strong maneuverability,short movement time,varied environment and poor regularity.It is difficult to capture,track and measure the position,attitude and speed of high-speed moving targets.Compared to traditional lidar tracking methods,vision-based detection and tracking can capture more motion information.With the development of high-speed vision sensors,it is increasingly important to use high-speed cameras to detect and track high-speed motion targets.However,the structure of the existing target detection and tracking algorithms is complicated,and it is difficult to meet the real-time requirements of high-speed vision systems.In order to detect and track the target of high-speed motion,this paper designs and builds a high-speed vision system,and based on the system,proposes a fast and effective target detection and tracking algorithm.The main research contents and contributions of this paper are as follows:(1)Designed to build a high-speed vision system of 200fps@640×480.Analyze the work responsibilities and working characteristics of each component of the system,combined with the motion characteristics of high-speed moving targets,select appropriate high-speed cameras,image acquisition cards and apertures,and increase lighting equipment to improve image quality.The high-speed continuous image acquisition experiments verify the rationality of the system design and the correctness of camera parameter settings.(2)Combining the advantages and disadvantages of interframe difference method and background subtraction method,a fast and effective target detection algorithm is proposed.In order to reduce the interference of the algorithm on illumination changes,the Cb component with better target clustering characteristics is extracted in YCbCr space.The motion region of the target is obtained by using the interframe difference method and projection technique.After the interference region is removed and the same target region is merged,the background subtraction method is used to accurately detect the target in the region,and the moving average method is used to update background in non-sports areas.In order to reduce the time consumption,the improved fast contour search method around eight neighborhood is used to search the target contour,and the ellipse fitting method is used to obtain a more accurate target center position for the motion blur targets.(3)In order to solve the hysteresis of the tracking algorithm,a target tracking algorithm based on prediction is proposed.By comparing the prediction tracking algorithm based on Kalman filtering and the prediction tracking algorithm based on trajectory fitting,the comprehensive prediction algorithm combining 3-point linear predictor and 5-point square predictor weighted combination is used to track and predict the target center,and the ROI is updated according to the predicted position.When the target is reproduced,the target detection is performed in the predicted ROI,reducing the search time.(4)By detecting and tracking the table tennis in different situations,the verification system algorithm can track the target with the speed of 4m/s at 1.3m distance in real-time,and the tracking error is no more than 3 pixels.The robustness of the algorithm to occlusion and similarly interfering objects is also verified.In the case where the ROI is used to reduce the detection range,the average running time of the system algorithm is 3.2 ms/frame.