期刊论文

中文

电讯技术

1001-893X

2018

58

9

1079-1085

10.3969/j.issn.1001-893x.2018.09.015

国家自然科学基金资助项目(61772561) 湖南省重点研发计划资助项目(2018NK2012) 中南林业科技大学研究生科技创新基金资助项目(20183027)。

本校为第一机构

针对传统Harris角点检测效率低、非极大值引起的伪角点多等问题,提出了一种自适应阈值和归一化互相关(NCC)与随机抽样一致算法(RANSAC)相结合的Harris图像匹配算法。首先,采用自适应方式抑制非极大值的方法对角点进行预筛选;其次,采用Forstner算子对角点进行二次筛选;接着采用归一化互相关匹配算法对检测的Harris角点进行粗匹配;最后采用随机抽样一致算法对图像进行精确匹配。实验结果证明改进的方法不仅缩短了角点检测和图像匹配时间,而且能有效提高图像的匹配精度。

Aiming at the inefficiency of traditional Harris corner detection and the problem of false corners caused by non-maximum values,a Harris image matching method combing adaptive threshold and Normalized Cross Correlation(NCC)and Random Sample Consensus(RANSAC)algorithm is proposed.Firstly,the method of non-maximum suppressing in adaptive mode is used to pre-filter the corner points.Secondly,the Forstner algorithm is used to further refine the corner points so as to improve the corner accuracy.Then,the NCC is used to coarse match the detected Harris corners.Finally,the RANSAC is used to precisely...