Abstract: Image matching is a key technology in the field of computer vision.An image matching technology based on imaging optical path is proposed to solve the problem of heavy computation，slow speed and low accuracy in image matching.Firstly，a camera model based on binocular vision is established，and the geometric model of the camera intersection field is determined according to the camera model’s imaging optical path.The largest possible image overlap region of the two images is estimated by means of geometrical relationship.Then the image features matching is carryed out in the regions by using the traditional features matching algorithm.Experimental results show that，the proposed method which compared with extracting and matching feature points in the whole image can effectively improve the matching speed and reduce the occurrence probability of mismatch.

Key words: image matching, imaging optical path, field intersection, image overlap

摘要： 图像匹配是计算机视觉领域的一个关键技术。针对图像匹配过程中特征提取运算量大、速度慢、匹配精度低的问题，提出一种基于成像光路的图像匹配技术。建立一种基于双目视觉的摄像机模型，根据该摄像机模型的成像光路确定摄像机视场交汇的几何模型，由几何关系估算出两幅图像中可能存在图像重叠的最大区域，再采用传统的特征匹配算法在该区域内实现图像的特征匹配。实验结果表明，相较于在整幅图像中提取特征点进行匹配，该方法能有效地提高匹配的速度，降低误匹配的发生几率。

关键词: 图像匹配, 成像光路, 视场交汇, 图像重叠