基于KAZE特征的露天矿无人机影像匹配

Matching of Open-pit Mine UAV Images Based on KAZE Features

无人机摄影测量是露天矿影像数据采集的重要方式。针对露天矿独特的阶梯状地形、常规的无人机影像匹配算法匹配点数量少、误匹配率高的问题,提出了基于KAZE特征进行露天矿无人机影像匹配。KAZE算法采用非线性尺度空间,能够更好地保留露天矿地物边缘信息。选取露天矿不同区域的5组无人机影像,分别与SIFT和SURF算法进行对比实验。实验结果表明,KAZE算法的正确匹配点数量和匹配精度优于SIFT和SURF算法,单点匹配时间较短,适用于露天矿无人机影像匹配。

UAV photogrammetry is an important method of image data acquisition for open-pit mines.In view of the unique step-shaped terrain of open-pit mines as well as the fewer matching points and high mismatch rates of conventional UAV image matching algorithms,a matching method for UAV images of open-pit mines based on KAZE features is proposed.The KAZE algorithm uses a non-linear scale space,which can better retain the edge information of open-pit mines.Five sets of UAV images from different areas of the open-pit mine are selected and the KAZE algorithm is compared with SIFT and SURF algorithm.The experimental results show that the KAZE algorithm is superior to the SIFT and SURF algorithm in terms of the number of correct matching points and matching accuracy,and the single point matching time is shorter.The proposed method is suitable for matching of UAV images of open-pit mines.