BRISK-DAISY无人机影像匹配算法研究

Research on BRISK-DAISY UAV image matching algorithm

影像匹配是无人机影像分析的关键步骤,通过对常用影像匹配算法的对比研究,提出一种基于BRISK特征点与DAISY描述子的影像匹配算法。该算法首先采用BRISK算法提取影像特征点,然后利用DAISY描述子进行特征描述,最后基于单映射变换矩阵的RANSAC算法进行影像精准匹配。文中对不同算法开展试验研究,对比分析影像匹配的特征提取数量、单个特征点提取时间、正确匹配对数和单次匹配平均时间、匹配正确率等5个指标。试验结果表明:与BRISK算法相比,单个特征点提取耗时约是BRISK算法的80%,单次匹配所需平均时间为BRISK算法的30%,甚至更短。与其它常规匹配算法相比,文中算法的匹配正确率与其他算法相当,但获得更多的正确匹配对数,并且文中算法在单次匹配平均时间、单个特征点提取时间与特征点提取数等方面更优。

Image matching is a key step in UAV image analysis.Based on the comparison of common image matching algorithms,this paper proposes an image matching algorithm based on BRISK feature points and DAISY descriptors.Firstly,the algorithm extracts image feature points by BRISK algorithm,then uses DAISY descriptors to describe the features.Finally,the RANSAC algorithm based on single mapping transformation matrix is used to accurately match images.In this paper,the experimental research on different algorithms is carried out,and the five indicators such as the number of feature extraction,the single feature point extraction time,the correct matching logarithm and the single matching average time,and the matching correct rate are compared and analyzed.The experimental result shows that compared with the BRISK algorithm,the single feature point extraction time consumption of the proposed algorithm is about 80%of the time-consuming BRISK algorithm,and the average time required for a single match is 30%or even shorter of the average time of the BRISK algorithm.Compared with other conventional matching algorithms,the correct matching rate of the proposed algorithm is comparable to other algorithms.More correct matching logarithms can be obtained.The average time of single matching,single feature point extraction time and feature point extraction are obtained,of which the numbers are better.