计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (23): 147-152.DOI: 10.3778/j.issn.1002-8331.1606-0072

• 图形图像处理 • 上一篇    下一篇

基于改进RANSAC算法的单应矩阵鲁棒估计方法

夏克付1,2,李鹏飞2,陈小平2   

  1. 1.安徽电子信息职业技术学院,安徽 蚌埠 233030
    2.中国科学技术大学 计算机科学与技术学院,合肥 230027
  • 出版日期:2017-12-01 发布日期:2017-12-14

Robust method for homography matrix estimation based on improved RANSAC algorithm

XIA Kefu1,2, LI Pengfei2, CHEN Xiaoping2   

  1. 1.Anhui Vocational College of Electronics & Information Technology, Bengbu, Anhui 233030, China
    2.School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027, China
  • Online:2017-12-01 Published:2017-12-14

摘要: 单应矩阵的鲁棒性和精度直接决定了其应用效果,如何利用RANSAC算法估计出鲁棒、精确的单应矩阵,仍是一个有待研究的热点问题。针对传统RANSAC算法迭代次数多、运行时间长、单应矩阵估计精度较低的问题,在SIFT特征匹配算法的基础上,从剔除样本集中不符合图像几何特性的部分外点、快速舍弃不合理单应矩阵和迭代精炼单应矩阵等方面对RANSAC算法进行改进,提出一种基于改进RANSAC算法的单应矩阵估计方法,提高了单应矩阵估计的精度和效率。实验结果表明,该方法有效解决了传统RANSAC算法存在的问题,能够快速、精确估计单应矩阵。另外,对于不同视角和大小的图像,该方法均具有较好的鲁棒性。

关键词: 单应矩阵, 估计, 尺度不变特征转换(SIFT)算法, 随机抽样一致性(RANSAC)算法, 特征匹配

Abstract: The robustness and precision of the homography matrix directly determine its application effect. How to use the RANSAC algorithm to estimate the robust and accurate homography matrix is still a hot issue to be studied. For the problem that the traditional RANSAC algorithm has many iterations, long running and low accuracy of homography matrix estimation, on the basis of SIFT feature matching algorithm, RANSAC algorithm is improved by eliminating some of the external points of the image, which is not in conformity with the geometrical characteristics of the image, and quickly discards the unreasonable homography matrix and iterative refinement of the homography matrix and so on, it presents a method for homography matrix estimation based on improved RANSAC algorithm, improves the accuracy and efficiency of the homography matrix estimation. The experimental results show that the method can effectively solve the problems of the traditional RANSAC algorithm, and can quickly and accurately estimate the homography matrix. In addition, the method has better robustness for different views and sizes of images.

Key words: homography matrix, estimation, Scale Invariant Feature Transform(SIFT), Random Sample Consensus(RANSAC), feature matching