基于近景影像的稠密匹配方法  

Dense Matching Method Based on Close Range Images

作　　者：刘文齐 孙立双[1] LIU Wenqi;SUN Lishuang(School of Geomatics and Transportation Engineering,Shenyang Jianzhu University,Shenyang 110000,China)

机构地区：[1]沈阳建筑大学测绘与交通工程学院,沈阳110000

出　　处：《河南科学》2023年第7期946-955,共10页Henan Science

基　　金：辽宁省教育厅科学研究项目(lnfw202013)。

摘　　要：针对三维重建中影像匹配所存在的同名点稀疏、分布不均和部分局部特征约束难以抵抗旋转、尺度和仿射变换所带来的负面影响,设计了基于近景影像的稠密匹配算法,首先利用ASIFT算法与RANSAC算法获取影像组稳定匹配点,构建Delaunay三角网,确定目标三角形内的像素集合,利用仿射变换将其映射至对应三角形内,结合极线约束与改进后的局部灰度特征约束进行验证.最终利用公开数据集与自建数据集进行算法的有效性验证,实验结果表明,算法匹配正确率均保持在93%以上,具备良好的匹配点加密效果;并在一定程度上解决同名点稀疏、分布不均等问题,同时算法本身对不同的影像变换表现出了较好的适应性与稳定性.In order to overcome the negative effects of rotation,scale and affine transformation caused by sparse and uneven distribution of homonymous points and partial local feature constraints in image matching in three-dimensional reconstruction,a dense matching algorithm based on close-range image is designed.First,stable matching points of image group are obtained by using ASIFT algorithm and RANSAC algorithm,and Delaunay triangulation network is constructed to determine the pixel set in the target triangle.It is mapped to the corresponding triangle by using affine transformation,and validated with epipolar constraint and improved local gray feature constraint.Finally,the validity of the algorithm is verified by using the public dataset and self-built dataset.The experimental results show that the matching accuracy of the algorithm in this paper is above 93%,and it has good matching point encryption effect.The algorithm can solve the problem of sparse and uneven distribution of homonymous points to a certain extent.At the same time,the algorithm itself shows good adaptability and stability for different image transformations.

关 键 词：近景影像 稠密匹配 仿射变换 极线约束 局部特征约束 

分 类 号：TP305[自动化与计算机技术—计算机系统结构]