面向板料精准堆垛的线特征位姿标定方法

Line Feature Pose Calibration Method for Accurate Sheet Stacking

中文关键词: 位姿估计  特征检测筛选  网格划分  机械臂抓取堆垛

英文关键词: pose estimation  feature detection and screening  grid division  mechanical arm grasping and stacking

中文摘要:

      目的 板料堆垛是现代流程工业生产中常见的生产环节，其堆垛质量影响着包装和运输安全。针对多边形板料自动堆垛中机械手抓取和对齐问题，提出一种基于网格筛选的线特征相机位姿标定方法，给出空间特征点的选择标准，解决因板料纹理不足带来的点特征精度下降的问题。方法 首先通过检测特征点来表示待匹配线，然后利用基于三角网格的线特征构型筛选算法，获得精确的匹配线;再利用单应性矩阵求解获得转换矩阵，最后通过转换在线板料图像，获得实际抓取点和转换角度。结果 通过桌面板料抓取实验和现场验证，重投影误差不超过2像素，1 m板料的堆垛误差为0.5 mm，结论 证明了此方法在板料自动堆垛过程中针对特征检测和位姿估计问题的有效性。

英文摘要:

      Sheet stacking is a common production link in the industrial production of modern process, and its stacking quality affects packaging and transportation safety. Aiming at the problem of robot arm grasping and alignment in automatic stacking of polygonal sheets, the work aims to propose a method for calibrating the pose of line feature camera based on mesh screening and put forward the selection criteria of spatial feature points to overcome the problem of point feature accuracy decline caused by insufficient sheet texture. Firstly, the line to be matched was represented by detecting feature points, and then the accurate matching line was obtained by the line feature configuration screening algorithm based on triangular mesh. Next, the transformation matrix was obtained by solving homography matrix, and finally the actual grasping point and transformation angle were obtained by transforming the online sheet image. Through the desktop sheet grabbing experiment and field verification, the error of re-projection was less than 2 pixels, and the error of 1 meter sheet stacking was 0.5 mm. It is proved that this method is effective for feature detection and pose estimation in the process of automatic stacking of sheets.