APPLICATION OF THE THREE-DIMENSIONAL IMAGE EDGE FEATURE EXTRACTION IN MECHANICAL ARM POSITION GRASPING AND ATTITUDE DETECTION

• Published in: International Journal of Mechatronics & Applied Mechanics Vol. 1; no. 10; pp. 80 - 88
• Main Authors: Wu, Huihui, Zeng, Xianrong, Lai, Yanjun, Liang, Mingfeng
• Format: Journal Article
• Language: English
• Published: Bucharest Editura Cefin 2021
• Subjects: Accuracy; Algorithms; Attitudes; Calibration; Cameras; Coordinates; Curve fitting; Edge detection; Feature extraction; Grasping (robotics); Humanoid; Image analysis; Image processing; Interdisciplinary subjects; Iterative methods; Kinematics; Lie groups; Local optimization; Mechanical arms; Monte Carlo simulation; Motion control; Object recognition; Robot control; Robots; Screw theory; Three dimensional imaging; Trajectory control; Trajectory planning; Workspace
• ISSN: 2559-4397; 2559-6497
• DOI: 10.17683/ijomam/issue10/v2.9

In order to detect the position and attitude of the mechanical arm in the process of grasping, the kinematics model of the humanoid mechanical arm is established by using the screw theory. Based on the Monte Carlo method, the workspace point cloud image of the mechanical arm is obtained; the pixel coordinates of the image are extracted through the contour extraction algorithm of Halcon software. Combined with the hand-eye relationship, the position relationship between the camera and the mechanical arm base coordinate system is obtained. In the C++ environment, the secondary development of the mechanical arm is carried out to realize the basic functions of object positioning and robot control. Through the image analysis in MATLAB, it is observed that the extraction effect of workspace boundary points is good, and the curve fitting error is small, which lays the foundation for the follow-up trajectory planning and motion control of humanoid mechanical arm. Through real-time data monitoring and data initialization after each task, the problems of local optimization and convergence stagnation can be well solved, so that the image processing results that meet the requirements of mechanical arm self-recognition can be obtained, the iterative ergodic efficiency can be better improved, and the subsequent processing time can be shortened. The simulation results show that the algorithm can get better edge detection results. Finally, the robot can grasp the object. The application of three-dimensional image edge feature extraction technology makes the selection of objects faster and more accurate, which makes up for the shortcomings of traditional grasping technology. After this research, the application of three-dimensional image technology to further improve the mechanical arm's ability to grasp objects will gradually become widespread.