测绘通报 ›› 2023, Vol. 0 ›› Issue (2): 91-96,103.doi: 10.13474/j.cnki.11-2246.2023.0046

• 学术研究 • 上一篇    下一篇

改进AKAZE算法的泥石流区无人机影像特征匹配

宗慧琳1,2, 袁希平2,3, 甘淑1,2, 张晓伦1, 梁昌献1, 赵振峰1   

  1. 1. 昆明理工大学国土资源工程学院, 云南 昆明 650093;
    2. 云南省高校高原山地空间信息测绘 技术应用工程研究中心, 云南 昆明 650093;
    3. 滇西应用技术大学, 云南 大理 671006
  • 收稿日期:2022-03-02 发布日期:2023-03-01
  • 通讯作者: 袁希平。E-mail:kmustyxp@126.com
  • 作者简介:宗慧琳(1987-),女,博士生,工程师,主要从事低空无人机摄影测量与遥感技术应用工作。E-mail:263286908@qq.com
  • 基金资助:
    国家自然科学基金(41861054)

An improved AKAZE algorithm for UAV image feature matching in debris flow area

ZONG Huilin1,2, YUAN Xiping2,3, GAN Shu1,2, ZHANG Xiaolun1, LIANG Changxian1, ZHAO Zhenfeng1   

  1. 1. Faculty of Land Resource Engineering, Kunming University of Science and Technology, Kunming 650093, China;
    2. Plateau Mountain Spatial Information Survey Technique Application Engineering Research Center at Yunnan Provinces Universities, Kunming 650093, China;
    3. West Yunnan University of Applied Sciences, Dali 671006, China
  • Received:2022-03-02 Published:2023-03-01

摘要: 针对在泥石流区灾害应急中使用无人机高分辨率影像特征匹配时时效性较低的问题,本文提出了一种改进AKAZE无人机影像特征匹配的算法。该算法首先使用AKAZE特征点检测算法提取局部稳定不变特征,用二进制描述符BEBLID描述检测到的特征点,采用最近邻次近邻距离比(NNDR)完成初步匹配;然后采用核线几何约束计算变换矩阵,达到内点提纯、提高匹配质量的目的;最后选取5组同一无人机序列影像进行特征匹配试验,分别与经典SIFT算法、AKAZE 算法、ORB算法进行比较。试验结果表明,该方法的匹配准确率与SIFT算法接近,略高于AKAZE算法,明显优于ORB算法,计算速度明显优于SIFT算法和AKAZE算法,基本达到ORB算法的计算效率。本文方法能较好地应用于对匹配精度和匹配时效均要求较高的泥石流场景无人机影像数据处理中。

关键词: 泥石流区无人机影像, 特征提取, BEBLID描述符, 核线约束, 影像匹配

Abstract: To solve the problem of low timeliness when UAV high-resolution image feature matching is used in debris flow disasters, an improved AKAZE UAV image feature matching algorithm is proposed in this paper. The AKAZE feature point detection algorithm is used to extract locally stable invariant features, and the binary descriptor BEBLID is used to describe the detected feature points. Then, the nearest neighbor distance ratio (NNDR) is used for preliminary matching. Furthermore, epipolar geometric constraint is used to calculate the transformation matrix to purify the inner points and improve matching quality. Five groups of UAV sequences images are selected for the feature-matching experiment and compared with the classic SIFT algorithm, AKAZE algorithm, and ORB algorithm. The experimental results show that the matching accuracy of the proposed method is close to that of SIFT algorithm, slightly higher than that of the AKAZE algorithm, significantly better than that of the ORB algorithm, and the calculation speed is significantly better than that of SIFT algorithm and AKAZE algorithm, basically reaching the calculation efficiency of ORB algorithm. The method proposed in this paper can be applied to the UAV image data processing in the debris flow scenes which requires high matching accuracy and matching time.

Key words: UAV images in debris flow area, feature extraction, boosted efficient binary local image descriptor, epipolar geometric constraint, image matching

中图分类号: