摘要
在结构复杂、纹理重复的场景中,现有的特征线匹配方法存在的误匹配率高的问题。为了解决上述问题,融合特征线区域信息和特征线局部特征点信息,提出了一种基于多重约束的特征线匹配方法。该方法首先采用LSD方法检测图像的特征线,利用SIFT方法检测图像特征点。融合特征线和特征点构建图像特征集,并计算特征线支持区域;其次在特征线支持区域内,依据特征线两端点及两侧紧邻点间的NCC度量值和核线约束,获得待匹配特征线的支持区域;然后从特征线间的角度、垂直距离和水平距离3个约束出发,采用特征线间的灰度相似性度量,在待匹配特征线的支持区域内进一步约束确定匹配特征线,从而有效解决了特征线误匹配问题。以不同场景的图像对为对象,实验结果表明所提出的方法在匹配数量上、匹配正确率上均优于其他方法,进一步验证了算法的有效性。
In the scene with complex structure and repeated texture,the existing image feature line matching methods have poor performance of high false matching rate.To improve the efficiency of feature line matching,a feature line matching method based on multiple constraints was proposed by fusing the feature line region information and the local feature point information.Firstly,LSD method was used to detect the feature lines of the image,and SIFT method was used to detect the feature points of the image.The feature line support region was calculated by the feature set constructing by fusing the feature lines and feature points.Secondly,the support region of the candidate feature line matching set was obtained according to the NCC measurement between the two ends of the feature line and the adjacent points on both sides and the core line constraint.Finally,from the perspective of angle,vertical distance and horizontal distance,the gray similarity measure between feature lines was used to determine the line to be matchedin the support region,so the mismatches of feature line were effectively solved.Taking the image pairs of different scenes as the dataset,the experimental results show that the proposed method is superior to other methods in terms of matching quantity and matching accuracy,and the effectiveness of the algorithm is further verified.
作者
王欣波
王亚伟
厉宝山
张宁
姜涛
徐刚
魏丽
Wang Xinbo;Wang Yawei;Li Baoshan;Zhang Ning;Jiang Tao;Xu Gang;Wei Li(Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China;Foshan Science and Automation Intelligence Technology Corporations,Foshan,Guangdong 528010,China;Bohai Shipyard Group,Huludao,Liaoning 125000,China)
出处
《机电工程技术》
2022年第9期76-79,共4页
Mechanical & Electrical Engineering Technology
基金
佛山市科技创新项目(编号:2017IT100022)。
