Research On Target Tracking And Positioning System Of Broken Robot Based On The Mobile Vision

The target tracking and positioning of mobile robots has become a research hotspot in the field of robotics.But at present,the research on the tracking and positioning of broken and dismantling robot is still in blank.It is easy for manual breaking to give rise to gravel spatter and collapse,and further cause secondary disasters.Therefore,it is significantly meaningful to carry out the study of the target tracking and positioning system of breaking robots to equip the breaking robots with the ability of tracking and positioning the breaking targets,moving to the target automatically,adjusting the breaking arm position and conducting breaking task.Through analysis of target tracking and positioning principles of breaking robots,it is known that: the key point to enable breaking robots to position and break targets automatically is to avoid losing the target and keep the camera on the right track when breaking robot is on the move.Therefore,this subject studies the real-time tracking methods of breaking robots towards the target during the move.Firstly,target tracking and positioning principles of breaking robots were analyzed to make sure that the camera conducted real-time tracking and kept itself on the right track when the breaking robot was moving.It was also the subject requirement to keep the target at the image center.Overall scheme of realizing the target tracking of breaking robots was designed: interest target was selected on the image captured by the camera;target tracking algorithm was operated to obtain the coordinate of the target center in the image;difference between the coordinates of target center and image center was calculated and converted into the rotation angle of steering cradle;steering cradle was controlled to rotate and drive the camera to be on the right tract so as to keep the target center at the image center and realize the target tracking.In the end,system platform composed of robots and target tracking module was established according to the overall scheme.Secondly,target features were selected and target tracking algorithm was determined.Feasibility of different target features was compared and color was chosen as the target feature in combination with the actual situation of breaking tasks.The Meanshift algorithm,Camshift algorithm and sequential monte carlo algorithm are compared together.Sequential Monte Carlo algorithm was improved combining Camshift algorithm.In the end,experiments were carried out to verify different types of algorithms.Thirdly,in order to make sure that the tracking algorithm was real-time and accurate so as to solve the wrong tracking of Camshift algorithm when the target was shaded,Camshift algorithm in combination with improved Sequential Monte Carlo algorithm was proposed.Bhattacharyya coefficient was taken to decide whether the target was shaded.When Bhattacharyya coefficient was greater than the threshold,it was believed that the target was not shaded,and Camshift tracking algorithm with good real-time property was operated correspondingly.When Bhattacharyya coefficient was smaller than the threshold,it was believed that the target was shaded,and improved Sequential Monte Carlo algorithm with good accuracy was selected to track the target.Fourthly,visual coordinate system was defined.Camera was calibrated with Zhang Zhengyou calibration method,with the parameter matrix inside the camera being calculated to deduce the conversion relations among target center,image center and rotation angle of the cradle.In order to prevent the frequent swing of the cradle,valid tracking range was set.When the target center was at the valid tracking range,it was believed that the tracking was successful.Arduino steering cradle control program was designed to realize the tracking of camera towards targets.Lastly,target tracking experiment was completed on the basis of the platform constructed in the former stage when the breaking robot was on the move and their tracking situations of interest targets in a state of moving forward,backward,turning left and right inside the indoor and outdoor environments are verified.Besides,the error analysis is done.The experimental results provide support for breaking robot navigation in the latter stage.