摘要
针对现在基于图像的三维重建面临的图像数据尺度不一和影像分辨率增高特征点增加两个问题,本文提出了基于SIFT的尺度均衡和均匀限制特征点数量的改进策略,通过采用给定特征点数量的方法来控制影像提取特征点的数量。同时根据高斯尺度增大与提取特征点数量成反比的特性,通过构建尺度与特征点数量的比例系数和给定特征点数量就可以获得影像的不同高斯尺度上提取的特征点数量。通过对影像进行格网划分,将提取特征点数量分配到每个格网。通过控制格网提取特征点数量的控制,可以在控制特征点提取数量的同时,获得更为均匀的特征点分布。实验证明,本文改进算法与SIFT算法相比,尺度分布更加均匀,空间分布的比例系数有30%左右的提升,提取特征点匹配的正确率有5%的提升,因此,本文改进算法在基于图像的三维重建中有一定程度的应用价值。
Aiming at the two problems of different image data scales and the increase of feature points with higher image resolution faced by image-based 3D reconstruction,an improved strategy based on SIFT-based scale equalization and uniformly limiting the number of feature points is proposed.In this paper,the number of feature points extracted from the image is controlled by using the method of given number of feature points.At the same time,according to the characteristic that the increase of Gaussian scale is inversely proportional to the number of extracted feature points,the number of feature points extracted on different Gaussian scales of the image can be obtained by constructing the scale factor and the number of feature points and the given number of feature points.By dividing the image into grids,the number of extracted feature points is allocated to each grid.By controlling the number of feature points extracted by the grid,a more uniform feature point distribution can be obtained while controlling the number of feature points extracted.Experiments have shown that compared with the SIFT algorithm,the improved algorithm in this paper has a more uniform scale distribution,a 30%improvement in the proportional coefficient of the spatial distribution,and a 5%improvement in the matching accuracy of the extracted feature points.Therefore,the improved algorithm in this paper is based on.There is a certain degree of application value in the three-dimensional reconstruction of the image.
作者
彭得阳
邓安健
PENG Deyang;DENG Anjian(School of Surveying and Land Information Engineering,He′nan Polytechnic University,Jiaozuo 454000,China)
出处
《测绘与空间地理信息》
2021年第12期46-50,56,共6页
Geomatics & Spatial Information Technology
基金
河南省自然科学基金面上项目(182300410113)
河南理工大学博士基金项目(B2017-08)资助。
关键词
SIFT
尺度分布
空间分布
信息熵
三维重建
SIFT
scale distribution
spatial distribution
information entropy
three-dimensional reconstruction
