工件检测中的图像模板匹配

　　针对在线工件的特点提出了一种新的工件检测办法。该算法提取方向码作为特征，分粗精两个步骤进行检测，首先计算模板和子图方向码并计算出方向码直方图，粗略的计算候选匹配点和工件旋转角度，然后，针对匹配候选点，利用子图和模板的方向码差值进行精匹配。该方法在工件位置发生旋转甚至是在遮挡和光强引起的光照波动等情况下，也能较好的进行匹配。

　　Abstract:

　　In this paper， a new approach to the workpiece detection is proposed according to the characteristics of online work. The algorithm extracts orientation codes as a feature and consists of two stages: rough test and precision test. First the matching candidates are selected using histograms of orientation codes and rotation angles are computed roughly. In the second stage， matching based on differences of orientation codes between template and rotated subimage is performed only for the matching candidates. Experiments demonstrate the proposed method can also better match even in the shelter and fluctuations in light.