• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 图形与图像 • 上一篇    下一篇

基于改进AKAZE特征的无人机影像拼接算法研究

宋伟1,王永波1,张培佩1,2   

  1. (1.中国矿业大学环境与测绘学院,江苏 徐州 221116;2.中国地质大学(武汉)地球物理与空间信息学院,湖北 武汉 430074)
  • 收稿日期:2018-05-17 修回日期:2018-09-26 出版日期:2019-05-25 发布日期:2019-05-25
  • 基金资助:

    国家自然科学基金(41271444)

An improved UAV image stitching
algorithm based on AKAZE feature

SONG Wei1,WANG Yongbo1,ZHANG Peipei1,2   

  1. (1.School of Environment Science and Spatial Informatics,China University of Mining and Technology,Xuzhou 221116;
    2.Institute of Geophysics & Geomatics,China University of Geosciences,Wuhan 430074,China)
  • Received:2018-05-17 Revised:2018-09-26 Online:2019-05-25 Published:2019-05-25

摘要:

针对目前无人机影像拼接时间效率较低,效果较差的问题,提出一种改进的AKAZE特征的无人机影像拼接算法。算法在影像匹配阶段利用BRISK描述符来代替MLDB描述符;采用特征点尺度信息和均方根误差作为约束条件剔除错误匹配;另外,将并行运算的思想应用于特征点提取和特征描述符的计算中来提升算法的时间效率。在影像融合阶段,利用RANSAC算法与LM算法相结合进行单应矩阵计算,以提高其计算精度,并采用多频段融合算法实现影像拼接。实验表明:通过改进算法在有效提升时间效率的同时,可以获得较高准确率和较高精度的匹配结果,实现无人机影像的快速无缝拼接。

关键词: 影像拼接, AKAZE, 尺度信息, 均方根误差, OpenMP, LM算法

Abstract:

Aiming at the low time efficiency and poor effect of UAV image stitching, we propose an improved  UAV image stitching algorithm based on AKAZE feature. In image matching phase, the MLDB descriptor is replaced by the BRISK descriptor. The scale information of feature points and root mean square error (RMSE) are taken as the constraint conditions to eliminate false matching points. In addition, the idea of parallel computation is applied to feature point extraction and feature descriptor calculation to improve the time efficiency of the algorithm. In image fusion stage, the RANSAC algorithm combined with the LM algorithm is used to calculate the monotone matrix so as to improve its precision. Finally, the multiband fusion algorithm is used to achieve image stitching. Experimental results show that the proposed algorithm can achieve more accurate and more precise matching results while improving the time efficiency, thus realizing fast and seamless stitching of UAV images.
 

Key words: image stitching;AKAZE;scale Information, root mean square error (RMSE);OpenMP, LM algorithm