摘要
ORB算法图像匹配效果很大程度上依赖于特征点描述子的质量,针对传统ORB特征描述子稳定性差、抗噪性弱等不足,提出一种改进的ORB特征描述子图像匹配算法。通过改进的动态自适应阈值FAST算法对图像进行特征点检测提取,并用改进的ORB特征描述子方法对待匹配的特征点进行描述。采用暴力匹配方法得到图像特征点粗匹配,最后利用RANSAC剔除误匹配特征点,筛选出精确匹配点。对比随机描述子和样本训练描述子方法实验表明,改进算法可有效提高图像匹配质量,在特征点匹配数目和正确率等指标上更优,验证了改进算法的有效性。
The image matching effect of ORB algorithm largely depends on the quality of feature point descriptors.Aiming at the disadvantages of traditional ORB feature descriptors,such as poor stability and weak noise resistance,an improved image matching algorithm for ORB feature descriptors is proposed.The improved dynamic adaptive threshold FAST algorithm is used to detect and extract image feature points,and the improved ORB feature descriptor is used to describe the matching feature points.The rough feature points of matching images are obtained by brute force matching method,and the incorrect matching feature points are eliminated by RANSAC algorithm to sift out the exact matching points.Compared with random descriptors and sample training descriptors,the experiment shows that the improved algorithm can effectively improve the quality of image matching,and it is better in the number and accuracy of matching feature points,which verifies the effectiveness of the improved algorithm.
作者
汪永生
李岩
刘明
WANG Yong-sheng;LI Yan;LIU Ming(School of Mathematics and Computing,Tongling University,Tongling Anhui 244061)
出处
《巢湖学院学报》
2020年第6期77-85,共9页
Journal of Chaohu University
基金
安徽省高校自然科学研究重点项目(项目编号:KJ2018A0474、KJ2017A467)。
作者简介
汪永生(1976-),男,安徽枞阳人,铜陵学院数学与计算机学院实验师,主要从事图形图像处理、增强现实技术研究。
