改进ORB算法在机器人分拣系统中的应用研究

Research on the Application of Improved ORB Algorithm in Robotic Sorting System

为满足工业上对表面具有复杂纹理特征工件的快速识别和分拣要求,将机器视觉引导技术应用到机器人分拣系统之中,提出了一种改进ORB算法,该算法采用oFAST特征提取和SURF特征子描述,完成对工件的快速识别,通过对工件形心坐标的计算,引导机器人进行分拣工作。结果表明:改进ORB算法与ORB算法相比,对尺度变化下工件图像的匹配准确率由59.52%提升到90.12%,对光照变换及噪声下的工件图像识别效果更好,将其应用到机器人分拣系统中可以对表面具有复杂纹理特征的工件进行准确识别与分拣,具有较高的鲁棒性。

To meet the industrial requirements for fast recognition and sorting of workpieces with com⁃plex texture features on the surfaces,an improved ORB algorithm is proposed by applying machine vi⁃sion guidance technology in a robot sorting system,which uses oFAST feature extraction and SURF fea⁃ture sub-description to complete fast recognition of workpieces and guides the robot to perform sorting work by calculating the workpiece shape center coordinates.The experimental results show that the im⁃proved ORB algorithm increases the matching accuracy of workpiece images from 59.52%to 90.12%under scale change.Compared with the ORB algorithm,and the recognition of workpiece images under light transformation and noise is better,and its application to the robot sorting system can accurately recognize and sort workpieces with complex texture features on the surface,with high robustness.