Research On Vision-based Target Detection And Tracking Of High-Speed Moving Objects

High speed moving object has the characteristics of fast speed and nonlinear motion.It has a wide range of applications for tracking high-speed moving objects,such as biomedicine,energy and chemical industry,military target detection and interception,etc.Although many scholars have made some achievements in the field of high-speed moving object tracking.CAMSHIFT is an improved algorithm of meanshift,which solves the problem that the tracking frame cannot be changed.At the same time,it uses HSV instead of RGB to solve the problem of light interference to a certain extent.But for the tracking of high-speed moving objects,using CAMSHIFT algorithm will encounter two problems.One is that CAMSHIFT only uses color histogram.For high-speed moving objects such as UAV or racing car,when the target rotates,the tracking effect of the target becomes worse;Second,in practical application,the target will be occluded,which affects the tracking accuracy.In order to solve the above problems,the main research contents of this paper are as follows:(1)This paper studies the method of target detection and tracking of high-speed moving objects,and designs a system of target detection,tracking and ranging.The system includes video stream processing,monocular ranging,target detection,target tracking and other modules,and realizes the tracking and ranging of high-speed moving objects.(2)Aiming at the high-speed sports car that is close to the camera,a similar triangle ranging algorithm based on the camera pose angle is proposed.By reducing the influence of the camera's attitude angle on the monocular ranging,and combining the image characteristics of high-speed moving objects,the problem of inaccurate ranging results for high-speed moving objects is solved.The experimental results show that the accuracy of the ranging method can be effectively improved by reducing the influence of the camera's attitude angle.(3)In order to solve the problem of light change when tracking high-speed moving objects,LBP feature and Haar feature are fused to obtain hlbp feature,which is used as the target matching feature.When the light changes,CAMSHIFT's color feature histogram is used to solve the problem of tracking failure;When the target rotates,the rotation invariance of hlbp is used to solve the problem of tracking failure,and the adaptive method is used to solve the weight problem of two features to achieve target tracking.Experimental results show that hlbp feature can effectively solve the problem of target rotation.(4)Aiming at the problem of occlusion in tracking high-speed moving objects,the Kalman filter method is used to predict the target,and the degree of occlusion is judged by the pasteurized distance,which solves the problem of tracking failure.The experimental results show that the fusion method of klaman filtering algorithm and CAMSHIFT algorithm can effectively improve the anti occlusion ability of the algorithm when the target is occluded in outdoor environment.(5)In the range finding and tracking system of high-speed moving objects,the improved CAMSHIFT algorithm is used to track UAV and racing car,and the algorithm is tested in real scene,which proves the effectiveness of the proposed method.