基于形状模板的快速高精度可靠图像匹配

Fast high-precision reliable image matching algorithm based on shape

为了提高工业检测中图像匹配精度和速度,提出一种基于形状模板的快速高精度图像配准算法:根据定义的图像匹配相似度量,采用图像金字塔搜索匹配策略,利用形状信息进行模板匹配。具体流程为:首先在参考图像上选择感兴趣区域生成模板,使用Canny滤波器对模板和搜索图像进行滤波,并计算边缘点的方向向量;其次,在此基础上构造该模板和搜索图像的图像金字塔,在图像金字塔最高层图像进行完全遍历匹配,获得具有匹配分值的潜在匹配点,然后根据匹配分值大小逐层逐次跟踪潜在匹配点,进行匹配,直至图像金字塔最底层;最后使用最小二乘法调整位姿参数,使其达到亚像素精度。实验表明该方法匹配速度快,匹配精度高,而且匹配鲁棒性高,不受遮挡、混乱、非线性光照变化、离焦、对比度低、全局对比度反转、局部对比度反转等情况的影响,完全可以满足实际工业需求。

A shape-based matching algorithm was proposed to improve the accuracy and speed of image matching in industrial detection. It is the template matching based on shape information, using image pyramid strategy, according to matching similarity measures defined. Specific process was as follows： first, the template was generated by selecting the interest region in the reference image. Then, the template and the search image were filtered using Canny operator, and the direction vectors of their edges were computed. The image pyramids were constructed for the filtered template and the filtered search image on this basis. The image matching was carried out on the highest level of the image pyramids according to the similarity measures defined. After the potential matches with matching scores were identified, they were tracked through the resolution hierarchy according to scores descending in successive until they were found on the lowest level of the image pyramids. Finally, the sub-pixel precision pose parameters were achieved through the least-squares method. Experimental results demonstrate that the proposed algorithm has fast speed and high accuracy, moreover it is robust to the occlusion, clutter, nonlinear illumination, defocus, low contrast, global contrast change, local contrast change, and so on, which meets the actual industrial demand effectively.