一种基于AKAZE算法的山区无人机影像匹配

Image matching of UAV based on AKAZE algorithm in mountain area

针对山区无人机影像地形起伏大、地貌特征复杂造成局部区域特征点探测难度大,匹配率不高的问题,文中通过改进算法,提高了山区无人机影像的特征匹配率。该算法首先采用AKAZE算法探测影像特征点,加入特征匹配多约束条件并基于单应性矩阵的RANSAC算法实现特征精匹配。同时为了验证改进算法的匹配精度,分别对SIFT、SURF两种算法进行试验对比分析,试验结果表明:针对山区无人机影像特点,通过改进算法能够较好地克服影像由于地形起伏大及地块特征破碎带来的影响,更好地剔除了特征点误匹配,匹配率更高。

It was difficult to detect feature points in local areas and the matching rate was not high by UAV images because of the large terrain undulations and the complex geomorphic features in mountainous.Aiming at these problems,the feature matching rate of mountain UAV images was improved by improving the algorithm.Firstly,AKAZE algorithm was used to detect image feature points,and the precise feature matching was achieved by adding the feature matching multiple constraints and the RANSAC algorithm based on the homography matrix.At the same time,in order to verify the matching accuracy of the improved algorithm,the SIFT algorithms and the SURF algorithms were compared and analyzed respectively.Experimental results showed that according to the characteristics of UAV images in mountainous areas,the improved algorithm could effectively overcome the impact caused by the large terrain undulations and the fragmented land features,and the feature point mismatching was eliminated,and the matching rate was higher.