| Intelligent robotic control can’t go without various forms of sensors, robots can be more aware of their surroundings and react properly by combing multiple sensors during the process of control. Since visual information has the advantage of wide coverage, rich information, etc., robot visual control has become a popular research direction in recent years. Machine vision based robot control is the result of multidisciplinary studies, in this paper, a Mitsubishi RV-4F robot visual control platform was established to achieve the goal of object pose estimation and grasp by analyzing robot kinematics, robot simulation, image processing and robot communication.In order to estimate the pose of an object from image, MATLAB was used to achieve image acquisition and image pre-processing. After that, by image binary segmentation, the target area was extracted from the background of the image and the object pose information in image could be obtained through geometric description of the area. Finally, according to camera perspective transformation and camera calibration can get the space coordinates of the object form its image coordinates.A semi-physical motion simulation platform was built in MATLAB for the purpose of visual information validation and real-time three dimensional (3D) motion simulation of industrial robot and its TCP/IP online control. For fast importing and real-time displaying with high accuracy of complex 3D models, a SPMD-based parallel processing method was proposed for rapid processing. The result proves of good performance.At last, a multifunctional robot simulation and visual control software was developed to achieve object location, movement simulation and robot visual grasping. Experimental results have shown that the platform is of good usability and high maneuverability, and can meet the needs of teaching and research work well. |
PDF Full Text Request