基于SURF和改进RANSAC的图像拼接方法  

Image stitching method based on SURF and improved RANSAC

作　　者：杨宇 赵成星 张晓玲 YANG Yu;ZHAO Chengxing;ZHANG Xiaoling(School of Electrical Information Engineering,Jiangsu University of Technology,Changzhou Jiangsu 213000,China)

机构地区：[1]江苏理工学院电气信息工程学院,江苏常州213001

出　　处：《激光杂志》2021年第4期105-108,共4页Laser Journal

基　　金：国家自然科学基金(No.61305123)。

摘　　要：在相机技术达不到超广角拍摄的背景下,图像拼接应运而生,从而得到视角更广阔的的图像。然而在进行图像拼接时候会存在较多的误匹配并且效率不高。针对误匹配,常常用RANSAC算法进行误匹配点剔除,但是在剔除过程中存在初始变换矩阵H的误差以及欧氏距离阈值设置的误差;针对匹配效率,由于RANSAC算法无上限迭代次数的原因,导致效率低下。提出一种RANSAC改进方法,在保留RANSAC方法基础上,引进HSI颜色模型模型相对应的色调、饱和度、亮度对匹配点进行约束。通过实验验证,此方法可以提高匹配的正确率,效率也明显得到改善,鲁棒性较好。There are many mismatches and inefficiency in image stitching.For mismatching,the RANSAC algorithm is commonly used to eliminate mismatching points,but there are errors in the initial transformation matrix H and errors in setting the Euclidean distance threshold;for matching efficiency,the RANSAC algorithm has no upper limit iterations,resulting in low efficiency.This paper proposes an improved RANSAC method and introduces the HSI model to constrain the matching points.This method can improve the accuracy of matching through experimental verification,and efficiency is also significantly improved.

关 键 词：SURF RANSAC HSI模型 图像拼接 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]