基于图像匹配和SURF的变压器姿态测量

Transformer posture estimation based on image matching and SURF

设计了一种以图像模板匹配和改进的具有鲁棒性的尺度不变特征算法(SURF算法)为基础的变压器姿态识别和角度检测算法,用于一种自动化装配线上小型变压器的检测和角度测量,配合机械臂对变压器的抓取,可实现在生产线上对变压器姿态调整。通过摄像头采集变压器图像,与模板进行匹配得到最优抓取变压器,利用SURF算法提取模板和匹配变压器的特征点,结合快速最近邻逼近搜索(FLANN)算法实现特征点的匹配,计算出匹配变压器相对于模板的旋转角度,从而确定变压器的摆放姿态。实验结果表明,该方案能够正确识别变压器图像,并对变压器相对于标准模板旋转角度进行准确测量,满足自动装配需求。

An algorithm based on image matching and SURF is designed for detection and angle measurement of a small transformer on an automated assembly line. It can cooperate with the manipulator arm to grasp transformers, and realize transformer posture adjustment on the production line. An isolated transformer can be selected after its image acquired by camera is matched with its template image. The feature points of the selected transformer image and the template are extracted respectively by the improved scale invariant feature algorithm （SURF）. Then the feature points matching are deduced by fast nearest neighbor search （FLANN） algorithm. Finally, the posture of transformer is calculated by the rotation angle of the transformer relative to the template. The experiment shows that the image matching of a transformer and its posture estimating can be obtained by the scheme, furthermore, the rotation angle of transformer respect to its standard template can be accurately measured, which meets the requirement of automatic matching.