计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (1): 292-299.DOI: 10.3778/j.issn.1002-8331.2007-0352

• 工程与应用 • 上一篇    下一篇

改进人工鱼群的ORB特征匹配算法

李思璇,胡志刚,王新征,付东辽,祖向阳   

  1. 1.河南科技大学 医学与技术工程学院,河南 洛阳 471003
    2.河南省智能康复医疗机器人工程研究中心,河南 洛阳 471003
  • 出版日期:2022-01-01 发布日期:2022-01-06

Improved AFSA for ORB Feature Matching Algorithm

LI Sixuan, HU Zhigang, WANG Xinzheng, FU Dongliao, ZU Xiangyang   

  1. 1.College of Medical Technology and Engineering, Henan University of Science and Technology, Luoyang, Henan 471003, China
    2.Intelligent Rehabilitation Medical Robot Engineering Research Center in Henan, Luoyang, Henan 471003, China
  • Online:2022-01-01 Published:2022-01-06

摘要: 针对室内轮椅定位与地图构建中传统ORB(oriented FAST and rotated BRIEF)受到特征点检测与选取策略的影响导致特征匹配正确率不理想,提出一种改进人工鱼群的ORB特征匹配算法。使用改进后的FAST检测特征点,利用改进后的人工鱼群在组合优化问题中具有收敛速度快且易获得最优解的特点,在图像中计算出不同特征区域,根据特征点所在区域位置赋予其相应的状态,对不同状态的特征点选择保留或去除,使用汉明距离的RANSAC算法在特征区域之间进行特征匹配。实验结果表明,改进后的FAST在图像边缘处提取到更多的图像特征,在实际环境中改进后的ORB匹配算法平均正确匹配率达到了92.7%,比传统ORB平均正确匹配率高52.3%。

关键词: 改进FAST, 改进人工鱼群, 提取与选取策略, 特征区域

Abstract: In the field of in indoor wheelchair localization and mapping , aiming at the problem of the unsatisfactory feature matching accuracy of the traditional ORB(oriented FAST and rotated BRIEF) with feature point detection and selection strategies, an ORB feature matching algorithm of improved artificial fish swarms is proposed. Firstly, the improved FAST is used to the feature points. Then, since the improved artificial fish population has the characteristics of fast convergence and easy to obtain the optimal solution in the combinatorial optimization problem, it is used to calculate different feature areas in the image, the corresponding state is given according to the location of the feature points, the feature points of different states are selected to be retained or removed, and finally the RANSAC algorithm of Hamming distance is used to perform feature matching between the feature regions. The experiments show that the improved FAST can extract more image features at the image edges, and the average correct matching rate of the improved ORB matching algorithm in the actual environment reaches 92.7%, which is 52.3% higher than the traditional average correct matching rate.

Key words: improved FAST, improved AFSA, extraction and selection strategy, feature areas