论文部分内容阅读
针对遥感影像匹配中经典SIFT算法提取特征点时,高斯分解易造成边缘模糊和纹理细节信息丢失,从而使得边缘匹配稳定性差和误匹配点多的问题,文章引入最新的基于非线性尺度空间的KAZE算法并提出了限以空间约束的方法进行多源遥感影像匹配:通过特征点优选并进行特征匹配来计算几何变换模型,并对匹配点的搜索空间进行约束来提高匹配速度和精度,最后通过均方根误差迭代剔除误配点。实验结果表明KAZE算法提取特征点比SIFT稳定性高,易于后期误配点剔除;限以空间约束的匹配策略优于传统匹配策略;对于细节及纹理模糊的影像,KAZE算法相比SIFT算法有独特的优势。
When the feature points are extracted from the classical SIFT algorithm in remote sensing image matching, Gaussian decomposition can easily lead to the loss of edge blurring and texture detail information, resulting in the problem of poor edge matching and mismatch points. This paper introduces the latest KAZE based on nonlinear scale space Algorithm and proposed a method of spatial constraint for multi-source remote sensing image matching: the geometric transformation model is calculated through feature point optimization and feature matching, and the matching point search space is constrained to improve the matching speed and accuracy. Finally, Square root error iteration to eliminate mismatch points. The experimental results show that the KAZE algorithm has higher stability than SIFT and is easy to eliminate mismatch points later. The matching strategy with space constraint is superior to the traditional matching strategy. For the detail and texture-blurred images, the KAZE algorithm has a unique Advantage.