Research And Implementation Of Mobile Robot Target Tracking Algorithm In Complex Scenes

With the continuous development of world economy and technology,mobile robots appear more and more frequently in human scientific research,production and life,and have broad application space.At the same time,with the development of computer and control technology,the application field of mobile robot is more and more extensive,and its working environment has also changed from a simple indoor environment to a variety of complex environments such as ground,underwater,air and even outer space,and target tracking research has gradually become a hot topic.Recently,the target tracking algorithm based on the depth neural network has been able to achieve high tracking accuracy and cope with a variety of complex environmental disturbances.However,the complexity of its algorithm model is too high to be real-time,which makes it temporarily stay in the stage of theoretical research and cannot be applied to the mobile robot platform.This topic aims at the environmental interference in the mobile robot platform to study the visual object tracking method suitable for carrying on it,and realize its application in mobile robots.The main research contents of this topic are as follows:Build the bottom layer of the mobile robot,install the four-wheel differential control method for the outer wheel,load the external sensor on the robot,build a real complex scene,use the image-based servo follow control strategy to achieve the motion control of the mobile robot,and identify and determine the target through the target detection algorithm of YOLOV4 and YOLOV5.Based on Deepsort tracking algorithm suitable for mobile robot platform,aiming at occlusion and interference of similar background on mobile robot platform,tracking result evaluation criteria are introduced to judge whether the target is subject to environmental interference,and motion target position estimation method designed by filtering is used to estimate the position of moving target subject to environmental interference,To solve the problem that the tracking algorithm can not well deal with occlusion and similar background interference,and improve the target loss relocation mechanism to increase the accuracy of target loss relocation.At the same time,introduce the tracking result evaluation standard of confidence to ensure the correct update of the target apparent model and provide a more accurate target model for target relocation,and improve the robustness of the visual tracking algorithm.On the basis of Deepsort algorithm,the trajectory prediction is carried out,and the Kalman filter is improved.The mechanism of trajectory prediction is introduced.Realize real-time target location prediction.