Research On Vision-Based Object Tracking Method For Mobile Robot

In recent years,with the wide application of mobile robots in various fields,people have become more and more demanding on their intelligence and autonomous interaction with the environment.Then it is important to obtain the shape,position,pose and speed of the target of interest in real time during the operation.Object tracking can obtain the relevant information by real-time identification and tracking of the target of interest.And because the environmental information obtained by the visual sensor is rich and easy to deploy quickly,vision-based object tracking method has been widely studied.However,the complex and changeable environment on the mobile robot platform places higher requirements on the real-time and robustness of the visual tracking method.This paper deeply studies the re al-time and robust visual object tracking method suitable for the mobile robot platform,and implements a vision-based mobile robot object tracking system.The main work of this thesis is as follows:1.Aiming at the problems of occlusion and similar background interference in the mobile robot's movement process,based on the fast discriminative scale space tracking algorithm,the tracking result evaluation criteria is introduced to determine whether the target is subject to environmental interference.At the same time,the moving object position estimation method is designed using Kalman filtering to estimate the position of the moving target affected by the environment.This method solves the problem that the f DSST tracking algorithm cannot well deal with occlusion and similar background interference,and improves the tracking performance of the track ing algorithm.And it reached 33 frames per second.2.Aiming at the problem of target loss during mobile robot movement,the visual tracking algorithm in the first work is improved from the perspective of the visual tracking algorithm framework,and a lon g-term visual tracking algorithm based on correlation filtering is proposed.This algorithm introduces a target loss relocation mechanism based on the candidate target position of Edge Boxes.It relocates targets in the case of target loss such as severe oc clusion and target out of view.It solves the problem that Kalman filtering cannot cope with the randomness of motion after target loss.It also uses different strategies to deal with different environmental interference problems.When the target is disturbed by the environment,the algorithm wil l first use the moving object position estimation method to predict the position of the target to filter out interference from similar backgrounds.Target relocation is only enabled when the tracking results for 30 consecutive frames are unreliable.At the same time,the algorithm also introduces high-confidence tracking result evaluation criteria to update the apparent model of the target more reliably and ensure the accuracy of the relocatio n results.It makes the algorithm robust in real time and has the capability of long-term tracking.3.Based on the research work of the visual object tracking algorithm,a visionbased mobile robot object tracking system was built,and the application verification of the visual tracking algorithm was implemented on the mobile robot pla tform based on the real scene.The error of the expected image coordinates and the actual image coordinates obtained by the visual tracking algorithm in real time is used as the input of the tracking control,and the image-based servo tracking control strategy is used to realize the motion control of the mobile robot,and then the visual tracking is realized.Aiming at the requirements of the mobile robot platform for the visual tracki ng method,this paper studies and introduces the tra cking result evaluation criteria on the basis of the efficient correlation filter tracking algorith m,and designs the moving object position estimation method and target loss relocation mechanism to deal with different environmental interference.This thesis takes the real-time as the premise to improve the robustness of the visual trac king method.Finally,this thesis designs and implements a vision-based mobile robot object tracking system,on which the verification of the visual tracking method is completed,and some basic research work is done for the vision-based mobile robot object tracking application.