Abstract: In the process of image feature extracting and matching，the mountain-climbing sequence is effective for solving the problems which is about the invariance to translation，scale and rotation，but the mountain-climbing sequence can not get a uniqueness of all the sequences.Focus on this condition，the improved algorithm based on the mountain-climbing sequence is presented and solves the nonuniqueness problem.The experimental results show that the improved algorithm based on mountain-climbing sequence decreases the max matching distance effectively than that of the original algorithm，also with the higher matching degree of similarity.

Key words: the mountain-climbing sequence, rotation invariance, uniqueness sequence, matching degree of similarity

摘要： 在图像特征提取和匹配过程中，爬山序列有效地解决了图像的平移、伸缩和旋转无关性问题，但是爬山序列可能存在旋转变换序列表现形式不唯一。针对此类情况，提出一种改进的爬山序列算法，能够较好地解决了图像旋转无关性问题。实验结果表明，相对于爬山序列算法，该算法有效地减小了平均最大匹配距离，从而使呈现出较高的匹配相似度。

关键词: 爬山序列, 旋转无关, 唯一性序列, 匹配相似度