Feature Recognition And Kinematic Inverse Solution Of Binocular Vision System For Industrial Robots

Industrial robots are the key technology to be developed in "Made in China 2025",and the development of machine vision technology has greatly expanded the robot’s ability to perceive the environment.With the increasing requirements of industrial robots in the field of intelligent manufacturing,it is of great value and practical significance to combine industrial robots with vision systems to build intelligent processing systems that can autonomously realize scene perception,target recognition,model construction,feature extraction and processing.However,industrial robots have complex nonlinear characteristics,and there are still many challenges in unifying the spatial perception capability of vision systems and the joint parameters of industrial robots to the same benchmark,and in reasonably modeling and accurately describing them.In view of the actual engineering applications,this paper takes industrial robots and machine vision in intelligent manufacturing systems as the research objects,and carries out research on key technologies such as modeling,performance analysis,hand-eye calibration,binocular point cloud alignment,point cloud pre-processing,feature recognition and inverse solution of industrial robot kinematics in combination with relevant experimental systems and simulation platforms for binocular vision systems.The main research contents are as follows:1.For the problems that the traditional algorithm has alignment accuracy to be improved for larger number of point clouds and is more sensitive to noise points in multi-vision systems,this paper proposes a composite point cloud alignment method based on improved whale optimization and improved ICP algorithm,which divides the point cloud alignment into two steps:initial alignment and accurate alignment.In the initial alignment stage,the random sampling consistency(RANSAC)alignment method is optimized by the improved whale optimization algorithm.In the early stage of the initial alignment,the global search capability of the whale optimization algorithm is enhanced by using a nonlinear convergence factor,and in the late stage of the initial alignment,adaptive weight coefficients are introduced to prevent the algorithm from rapid degradation and enhance the global expansion capability of the algorithm to avoid falling into local optimum.Through the initial alignment,the point cloud to be aligned is converted to the vicinity of the theoretical point cloud,and the rotation matrix and translation vector are calculated.In the exact alignment stage,the standard ICP algorithm is improved by introducing the key point normal vector weighting judgment,etc.The experimental results show that the method in this paper is robust and can achieve the exact alignment of complex point clouds,effectively avoiding the characteristics of ICP algorithm that are sensitive to outlier points and initial alignment poses.2.To address the problem that the traditional method has more feature loss after filtering in the point cloud pre-processing stage,an improved bilateral filtering algorithm based on neighborhood analysis is proposed to realize the pre-processing of point clouds.The experimental results show that the filtering method proposed in this paper can better deal with the noise points and outliers in the process of workpiece point cloud acquisition,and the overall features of the processed point cloud are better maintained,the boundary is more complete,and the number of point clouds is significantly reduced.The simulation and experimental data show that the proposed method can quickly and accurately identify the abnormal features of the curved workpiece point cloud and provide the workpiece information for the subsequent processing.3.To address the problem that the solution efficiency of the existing inverse kinematic solution of industrial robots needs to be further improved,an inverse solution solving method based on ELM-SSA-SCA is proposed.The method divides the inverse solution calculation into two steps.First,the initial inverse solution of the industrial robot is predicted by using an Extreme Learning Machine(ELM)with fast training speed.Then,in order to find the optimal inverse solution,a composite optimization method combining the improved Sparrow Search Algorithm(SSA)and Sine Cosine Algorithm(SCA)is proposed to optimize the obtained initial inverse solution,and a fitness function is designed to minimize the sum of the mean square error of the training and test sets,and the solution under the optimal fitness is taken as the optimal inverse solution.The method takes full advantage of the fast iterative capability of ELM and the faster speed of SSA seeking.In this paper,a 6-DOF industrial robot is used as an example,and its positive and inverse kinematic equations are established and verified based on the DenavitHartenberg(D-H)method.Then,the method is validated by 4000 sets of positional data.The experiments show that the algorithm designed in this paper has the advantages of fast convergence,strong optimization capability and smooth iteration compared with the traditional particle swarm-neural network algorithm(PSO-BP)based solution method.In summary,this paper investigates some key technologies of binocular vision system for industrial robots in intelligent manufacturing environment,and innovates in the fields of binocular point cloud alignment based on improved intelligent optimization algorithm,abnormal point cloud identification and inverse solution of industrial robot kinematics,etc.The proposed method has high accuracy and robustness,and the research results have certain theoretical significance and engineering value in the industrial field.