Target Tracking Method for Visual Autonomous Localization of Demolition Robot

To solve the problem of autonomous and precise positioning of the robot arm in the non-structural environment of the demolition robot, binocular vision is used for the measurement of the target point for robot arm positioning, and the real-time tracking and measurement of the end of the robot arm during the positioning process. For demolition robots operating in non-structural environments, their tracking process will have the problems of randomly appearing tracking targets being obscured and a complex background environment. The paper solves this problem by extending the Camshift algorithm to 3D space, using depth information to separate occluders and targets in space. Through experiments, it is proved that this method can effectively solve the problem that the tracking of the end of the robotic arm is lost due to random occlusions during the positioning and tracking of the end of the robotic arm, so as to provide the feedback information of the spatial position of the end of the robotic arm for the autonomous positioning of the robotic arm in a real-time and stable manner.