| 一种改进RANSAC算法的单应性矩阵估计方法 |
| |
| 引用本文： | 李嘉惠,张丰收,崔浩阳.一种改进RANSAC算法的单应性矩阵估计方法[J].教育技术导刊,2020,19(2):149-152. |
| |
| 作者姓名： | 李嘉惠 张丰收 崔浩阳 |
| |
| 作者单位： | 1.河南科技大学 机电工程学院，河南 洛阳 471000；2.上海大学 机电工程与自动化学院，上海 201900 |
| |
| 基金项目： | 国家重点专项（2017YFB0306401） |
| |
| 摘 要： | 在图像拼接技术中，单应性矩阵是实现两幅图像正确拼接的关键因素。针对传统RANSAC算法误匹配点概率较高，需要设置固定的投影误差阈值t导致迭代次数多、运行时间长、估计的单应性矩阵精度低等问题，提出一种改进的RANSAC算法以降低误匹配率。利用特征点周围灰度梯度相似性，剔除初始匹配中部分误匹配点，以减少矩阵估计的迭代次数；通过快速舍弃错误的单应性矩阵以减少内点检测时间，提高算法运行效率；通过BGD算法最小化损失函数以拟合精确的单应性矩阵。对比实验结果表明，改进的RANSAC算法能够有效剔除误匹配点，减少内点检测时间，提高单应性矩阵H的精度。
|
| 关 键 词： | 图像匹配 RANSAC算法 单应矩阵 BGD算法 误匹配点 |
| 收稿时间： | 2019-04-02 |
A Homography Matrix Estimation Method Based on Improved RANSAC Algorithm |
| |
| LI Jia-Hui,ZHANG Feng-shou,CUI Hao-Yang.A Homography Matrix Estimation Method Based on Improved RANSAC Algorithm[J].Introduction of Educational Technology,2020,19(2):149-152. |
| |
| Authors: | LI Jia-Hui ZHANG Feng-shou CUI Hao-Yang |
| |
| Institution: | 1.College of Mechanical and Electrical Engineering， Henan University of Science and Technology， Luoyang 471003， China；2.College of Mechanical and Electrical Engineering and Automation， Shanghai University， Shanghai 201900，China |
| |
| Abstract: | In image stitching technology， the homography matrix is the key factor to achieve the correct stitching of two images. The traditional RANSAC algorithm has a high probability of mismatching points and needs to set a fixed projection error threshold t and it will lead to many iterations， long running time， and low accuracy of the estimated homography matrix. An improved RANSAC algorithm is proposed to reduce the mismatch rate. The similarity of gray gradient around feature points is used to eliminate some mismatched points in initial matching， so as to reduce the iteration times of matrix estimation. By quickly discarding the wrong homography matrix， the detection time of the inner point is reduced， and the operating efficiency of the algorithm is improved. The loss function is minimized by the BGD algorithm to fit the exact homography matrix. According to the results of the comparative experiment， the improved RANSAC algorithm can effectively eliminate the mismatch points， reduce the inside point detection time， and improve the accuracy of the homography matrix H. |
| |
| Keywords: | image matching RANSAC algorithm homography matrix BGD algorithm mismatch point |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
