Research On 3D Reconstruction Technology Of Monocular Vision Based On Industrial Robot

Thanks to the promotion of "Made in China 2025",3D reconstruction technology has been widely used in various fields of manufacturing industry.Aiming at the problem of high cost and complex operation of 3D reconstruction technology in mechanical parts' assembling and inspecting,a low-cost,simple structure and flexible operation 3D reconstruction platform is proposed by using industrial robots and monocular cameras,in order to obtain a 3D reconstruction model with high detail reduction and accurate size.Firstly,in order to ensure the industrial robot can run smoothly around the reconstructed object and obtain a large number of its pictures while mounting the monocular camera,the industrial robot's motion analysis is carried out,the position,attitude description and coordinate transformation of the industrial robot are introduced,and the corresponding mathematical model is established to solve the forward and inverse kinematics.In the stage of trajectory planning,an improved trajectory optimization algorithm for industrial robots is proposed to ensure more stable motion of the robot.Secondly,the basic theories of pinhole camera model,coordinate transformation,polar constraints,basic matrix and eigenmatrix are introduced.SIFT algorithm is used to detect and match feature points after acquiring a large number of disordered target objects' pictures by camera,in order to ensure enough matching logarithms of edge points,SURF algorithm is used to detect and match the feature points again,and all the matching logarithm are used as the final result.Then introduce the sparse and dense points cloud reconstruction algorithm,elaborating on the algorithm principle and implementation process.Aiming at the problem that PMVS algorithm is difficult to remove outliers when reconstructing dense point clouds in small-scale scenes,a fusion improved point cloud filtering algorithm is proposed,using the algorithm we can divide a large number of outliers while ensuring the detail of target objects and the subsequent workload are reduced.Finally,a 3D reconstruction experimental platform is built to carry out trajectory optimization simulation experiments and dense point cloud filtering optimization experiments.The experiments show that the improved robot trajectory is more sommth and stable,which ensures the stability of industrial robot motion.The fused point cloud filtering algorithm removes a large number of outliers while guaranteeing the surface features of the reconstructed model,and the accuracy between the model and the actual measurement reaches 99%.