测绘通报 ›› 2020, Vol. 0 ›› Issue (6): 53-56.doi: 10.13474/j.cnki.11-2246.2020.0180

• 学术研究 • 上一篇    下一篇

利用A-AKAZE算法进行喀斯特地区无人机影像匹配

李闯1, 王晓红2, 何志伟3   

  1. 1. 贵州大学矿业学院, 贵州 贵阳 550025;
    2. 贵州大学林学院, 贵州 贵阳 550025;
    3. 贵州省测绘产品质量监督检验站, 贵州 贵阳 550004
  • 收稿日期:2019-09-23 修回日期:2019-12-17 出版日期:2020-06-25 发布日期:2020-07-01
  • 通讯作者: 王晓红。E-mail:gzdxwxh@163.com E-mail:gzdxwxh@163.com
  • 作者简介:李闯(1994-),男,硕士生,研究方向为无人机影像处理与应用。E-mail:lchuang1314@163.com
  • 基金资助:
    贵州省自然科学基金(黔科合J字[2014]2070);贵州省科技计划(黔科合LH字[2014]7649)

UAV image matching in Karst area using A-AKAZE algorithm

LI Chuang1, WANG Xiaohong2, HE Zhiwei3   

  1. 1. College of Mining, Guizhou University, Guiyang 550025, China;
    2. College of Forestry, Guizhou University, Guiyang 550025, China;
    3. Guizhou Province Surveying and Mapping Product Quality Supervision and Inspection Station, Guiyang 550004, China
  • Received:2019-09-23 Revised:2019-12-17 Online:2020-06-25 Published:2020-07-01

摘要: 喀斯特地区地形复杂,无人机影像匹配难度大、耗时多。针对如何提高该区域无人机影像的匹配效率,本文提出了一种基于AKAZE的改进算法。该算法首先利用完全仿射不变框架对原始影像进行视角模拟;然后利用AKAZE算法对模拟影像进行特征点提取和描述,并获得原始影像的特征点和描述符;最后利用基于单应性矩阵的RANSAC算法对原始影像进行精匹配,进而剔除粗匹配过程中错误匹配点对。本文对该改进算法开展了试验研究,并与ASIFT和AKAZE等常用算法进行了试验对比分析。试验结果表明,对喀斯特地区无人机影像匹配而言,与ASIFT算法相比,在保持相当匹配正确率的情况下,基于A-AKAZE算法的匹配总耗时是ASIFT算法耗时的50%左右,可以较大幅度地减少匹配总耗时;与AKAZE算法相比,基于A-AKAZE算法的影像总匹配对数及正确匹配对数至少是ASIFT算法的影像总匹配对数及正确匹配对数的7倍。综合考虑匹配耗时和正确匹配对数,本文算法优于AKAZE和ASIFT等常用算法,更适合于喀斯特地区的无人机影像匹配。

关键词: A-AKAZE算法, 影像匹配, 无人机影像, AKAZE算法, 喀斯特地区

Abstract: The terrain in Karst area is complex, and UAV image matching is difficult and time-consuming. Aiming at how to improve the matching efficiency of UAV images in this area, this paper proposes an improved algorithm based on AKAZE. The algorithm firstly uses the complete affine invariant framework to simulate the original image. Then the AKAZE algorithm is used to extract and describe the feature points of the simulated image, and obtain the feature points and descriptors of the original image. Finally, the original image is matched precisely by using the RANSAC algorithm based on homography matrix, and then eliminated error matching point pairs. In this paper, the improved algorithm is experimentally studied and compared with ASIFT and AKAZE. The experimental results show that, compared with ASIFT algorithm, the total matching time based on A-AKAZE algorithm is about 50% of the total matching time of ASIFT algorithm, and the total matching time can be greatly reduced under the same matching accuracy. Compared with AKAZE algorithm, the total matching logarithm and correct matching logarithm based on A-AKAZE algorithm are at least 7 times of the total matching logarithm and correct matching logarithm of ASIFT algorithm. Considering the matching time-consuming and the correct matching logarithm, this algorithm is better than AKAZE and ASIF, and more suitable for UAV image matching in Karst area.

Key words: A-AKAZE algorithm, image matching, UAV image, AKAZE algorithm, Karst area

中图分类号: