Research On Moving Target Algorithms Of Rotor-wing UAV Based On Vision

Vision-based targets tracking of UAV has the advantages of low cost and good effect,and is widely used in all walks of life and has a good development prospect.In the process of target tracking,the visual tracking algorithm is the core content,and the accuracy of the tracking algorithm determines the success or failure of the tracking.However,the complex scenarios that cannot be predicted during the tracking process seriously affect the accuracy of the tracking algorithm and need to be overcome by improving the algorithm,such as illumination changes,scale changes,occlusion,and background interference.This paper chooses to use the Mean Shift algorithm based on color feature tracking to improve it to solve the complex environmental problems faced by UAV when tracking targets,such as color interference and occlusion problems.The main research contents of this paper are as follows:(1)For the problem that the target color is similar to the environmental background when the target is tracked by UAV,the tracking feature of the Mean Shift algorithm is proposed.The video format captured by the camera is converted from the RGB color space to the HSV color space,and the hue component is selected.As a tracking feature,it is more sensitive to color than the RGB color histogram as a tracking feature,and is not susceptible to similar color interference.(2)Aiming at the problem that the target is short-term occlusion during the tracking process of the UAV,the improved Mean Shift algorithm is combined with Kalman filtering,and the prediction function of Kalman filtering is used to make the algorithm stable tracking in the face of short-term occlusion.(3)When using the Mean Shift algorithm for UAV tracking,the algorithm has the drawback of not being able to adaptively adjust the search box scale.In order to solve this problem,this paper proposes to add a binocular vision system to measure the distance between the tracking target and the UAV,thus controlling the distance between the two,ensuring that the tracking target will not be lost.Through experiments,it is verified that the tracking algorithm proposed in this paper can achieve better tracking effect in the face of complex scene tracking.