Research On 3D Perception And Anti-collision Technology Of Nuclear Decommissioning Robot Based On Multi Vision

The decommissioning of nuclear facilities is a worldwide issue.Nuclear decommissioned robots can replace operators and operate in high-radiation environments.Three-dimensional perception and collision avoidance operations are one of the important contents of nuclear decommissioned robots.When nuclear decommissioned robots are working,the operator observes the situation on the site through a monocular camera,and the human eyes recognize and locate the robot to manually operate the robot to complete the dismantling operation.Observed by the monocular camera,it is prone to occlusion.The operation of the operator is difficult and the efficiency is low.And the automation level of the operation process is not high.Therefore,in order to improve the efficiency of demolition,this paper uses the robot project under strong radiation field as the background to carry out research on the three-dimensional perception and obstacle avoidance technology based on multi-eye vision.The main research contents are as follows:First of all,this article analyzes the task requirements of the subject and clarifies the technical difficulties of the subject;compares and analyzes the three positioning methods of monocular vision three-dimensional positioning,binocular vision three-dimensional positioning,and multi-eye vision three-dimensional positioning.Combined with the requirements of the subject,it is finally determined to adopt multi-eye The vision system performs three-dimensional perception of the work scene,and then guides the robot to plan an overall technical solution for an obstacle-free and optimized path,and builds a multi-eye vision system.Secondly,the imaging model of the camera is introduced,the coordinate system of the vision system is established,and the transformation relationship of the projection matrix is deduced by combining with the distortion model;the vision system built in this paper is calibrated by Zhang Zhengyou’s calibration method and MATLAB software,and the distortion of the image is corrected,According to the actual working conditions,the corresponding algorithm is selected for preprocessing;according to the three situations of pipeline distribution,algorithms are designed to identify the pipeline to be dismantled.Then,perform stereo correction on the preprocessed image;introduce the stereo matching algorithm,compare the pros and cons of the algorithms,determine that the global binocular camera uses the BM algorithm,and the local binocular camera uses the SGBM algorithm,and the BM algorithm is improved to improve efficiency;Use the disparity value obtained by stereo matching and the camera calibration result to perform three-dimensional reconstruction of the target object.Thirdly,research on robot anti-collision technology.When the robot is operating manually,the KD-Tree-based point cloud anti-collision technology is used to set a safety threshold.When the distance between the robot and the obstacle point cloud is less than the threshold,the robot will stop automatically;when the robot works automatically,the improved RRT algorithm proposed in this paper is used for path planning,and simulated on MATLAB software.Through comparison and analysis with traditional RRT and RRT-connect path planning algorithms,the feasibility of the proposed algorithm is verified.Finally,a multi-eye vision-based 3D perception and obstacle avoidance experiment system for nuclear decommissioned robots is built,software and hardware platforms are designed respectively,and the built experimental platform is used to verify the robot obstacle avoidance operation effect.Experiments show that the system can successfully complete the threedimensional scene perception and obstacle avoidance tasks,which proves the feasibility of the research in this paper.