室内场景下目标的识别抓取算法研究

Research on Target Recognition and Grasping Algorithms in Indoor Scenes

针对室内环境下对小目标物体进行识别抓取的算法进行了研究。使用YOLOv5算法,基于RealSense D415深度相机对待抓取物体进行目标识别和分类,确定物体距离相机坐标系的距离信息,利用Canny、HoughLines线条检测提取物体在2D下的姿态信息并映射到3D中进行姿态判定;根据姿态,机器人末端手爪旋转相应的角度,实现机器人对目标物体的准确抓取。最终实验结果表明:该算法的运行速度为30fps,对水瓶以及花瓶可以成功抓取,在复杂的背景条件下也可以识别。

In this paper,the algorithm of recognizing and grasping small target objects in indoor environment is studied.It is proposed to use YOLOv5 algorithm,and to recognize and classify grasped objects based on RealSense D415 depth camera.The distance information of the object from the camera coordinate system is determined,and the attitude information in 2D extracted by Canny and HoughLines line detection is mapped to 3D for attitude determination.According to the attitude,the end claw of the robot rotates the corresponding angle,so that the robot can grasp the target object accurately.The final experimental results show that the running speed of the algorithm is 30fps,the water bottle and vase can be grasped successfully,and can be identified under complex background conditions.