基于LK光流的模板匹配算法的研究与改进

Analysis of Motion Planning on Bond Arm

针对半导体设备中机器视觉的识别定位问题,介绍了传统的LK光流的模板匹配算法,在此基础上提出了附加运动方向光滑约束条件的LK光流的模板匹配算法。该算法采用运动方向光滑约束条件作为正则化项,较好的保持了运动的边界。通过实验分析,采用该算法对芯片进行识别定位,具有精度高,计算速度快的特点。

In view of Machine Vision system, the thesis introduces template matching algorithm based on the optical flow of LK firstly.Then an optical flow method with the motion oriented smoothness constraint is proposed,which adoptes motion oriented smoothness as the restraint constraint.The experiments prove that this improved algorithm can enhance the accurateness of template matching effectively, Meanwhile can reduce the time of matching.