摘要
基于灰度相关的模板匹配算法在很多情况下难以得到准确的匹配结果,提出一种基于边缘几何特征的高精度模板匹配算法。利用曲面拟合方法获得边缘的梯度方向和亚像素坐标作为匹配信息,采用图像金字塔的搜索策略对算法加速,最后利用最小二乘平差理论得到亚像素级的定位精度和精确的旋转角度信息。实验表明,对于目标旋转、均匀或非均匀变化的光照、部分遮挡的情况下可以得到良好的匹配结果,而且在保证高精度的同时算法可以满足实时性要求,重复定位精度优于商业化软化包MIL8.0的GMF算法。
Gray-scale correlation based template matching algorithm can hardly obtain accurate matching results in some conditions, a fast and high precision template matching method based on edge geometric features is proposed. Surface fitting method is used to obtain the gradient directions and sub-pixel coordinates of the edge points, which are used as the matching information in calculating the similarity between template and target. In order to satisfy the real- time requirement, image pyramid searching strategy is employed to accelerate the algorithm. Furthermore, the least square adjustment theory is adopted to calculate the sub-pixel positioning precision and precise rotation angle infor- mation. Experiment results demonstrate that the algorithm introduced in this paper can obtain good matching results in the case of target rotation, uniform or noon-uniform illumination disturbance, partial occlusion, and etc. Moreover, besides the stability, reliability and high precision, the proposed algorithm also can meet the real-time requirements. The repeat positioning precision of the algorithm is better than that of the GMF algorithm of the commercialized machine vision package of MIL8. 0 from Matrox.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2013年第7期1462-1469,共8页
Chinese Journal of Scientific Instrument
基金
广东省科技计划(2012B091000020)
深圳市科技计划(JC201104210015A
CXB201105100073A)
深圳南山科技计划(201002)资助项目
关键词
模板匹配
边缘特征
曲面拟合
最小二乘平差
template matching
edge feature
surface fitting
least-square adjustment
