| With the development of artificial intelligence,service robots have been widely used in daily life.Traditional service robots often use offline programming or teaching methods to complete the grasping task.With the continuous development and improvement of computer vision technology,robots that introduce vision can complete more complex and diverse tasks.Grasping object recognition and posture estimation is one of the important areas of research on visual grabbing service robots.However,2D images can only recognize the position of objects,and cannot obtain depth information of objects.However,traditional 3D point cloud visual detection algorithms are complex and time-consuming.Therefore,this article has deeply studied the target detection and pose estimation algorithms in the field of computer vision.Using the object recognition and pose estimation methods combined with 2D color maps and 3D depth maps,a robotic arm grasping system based on computer vision has been designed to complete the recognition and pose estimation of target objects in complex environments.The robotic arm completes the grasping task based on its pose,improving the intelligence of the service robot.The main content of this article is as follows: Complete the hardware selection and platform construction according to the functional requirements of the visual grabbing service robot,and complete the overall design of the visual grabbing service robot system.Then,the visual grasping system is calibrated to obtain the conversion relationship between the camera coordinate system and the robot arm end coordinate system.For the target detection algorithm,the improved YOLOv7-tiny algorithm is used to achieve an average accuracy of 94.74% on the captured object dataset in this article.The detection and positioning frame for capturing the target object is obtained,and the color map and depth map are matched to initially determine the type and location of the target object.For pose estimation algorithms,the scene point cloud is segmented and processed using RANSAC,clustering,and other algorithms to obtain each target object.The target object type is determined using a target detection algorithm.The PCA pose estimation algorithm is used to obtain the pose of each captured object.The improved PCA algorithm is used for the box pose estimation,which reduces the number of point clouds by 8.9% compared to the traditional PCA pose estimation algorithm.A scoring strategy is developed to obtain a priority for capturing pose.The data communication between the camera and the robotic arm is established in the ROS operating system.The robotic arm grasps based on position and posture information to complete the grasping task in complex scenes.The success rate of single target grasping reaches 90.8%,and the success rate of multi target grasping reaches 75%,verifying the feasibility of the system scheme in this paper. |
