根据大型工件测量系统中的图像拼接特点，提出了一种改进的模板匹配与特征提取相结合的高精度快速图像拼接算法．该算法对相邻的两幅图像采用多尺度Harris 算子提取特征点，在第一幅图像中根据特征点的分布确定模板块，在第二幅图像中进行模板匹配找到匹配块，再基于特征提取在匹配块内进行对应特征点的计算，完成拼接．试验结果表明，该算法大大减小了图像拼接的计算量，并有效解决了工件表面特征稀少且存在大量相似区域、因而极易产生伪匹配的问题，能满足大型工件测量系统中的高精度快速图像拼接的要求．

According to the characteristics of the image mosaic of the large-scale workpiece measurement system,an improved fast mosaic algorithm with high accuracy is proposed by combining the feature extraction with the template matching. In this algorithm,first,the feature points of two adjacent images are extracted by using the multiscale Harris operator. Then,in the first image,the templates are identified according to the distribution of the feature
points; and in the second image,the matching blocks are chosen by means of the template matching. Finally,the matching feature points in the matching blocks are detected by the feature matching,and the mosaic is thus completed. Experimental results show that the proposed algorithm greatly reduces the computation amount of the mosaic,effectively avoids the false matching which is easy to occur due to scarce features and large-scale similar areas of the workpiece,thus meeting the requirements of the measurement system.