倾斜影像的特征提取及匹配研究

发布时间：2018-01-21 00:19

本文关键词： SIFT MSER 倾斜影像特征提取图像匹配　出处：《长安大学》2017年硕士论文　论文类型：学位论文

【摘要】：信息革命爆发以来,信息的存储、传输、表示方式发生了质的变化,直接导致地理信息在数据获取、处理与可视化显示等方面技术的飞速进步。倾斜摄影测量是航空摄影测量技术上的一大突破,通过在同一飞行平台上搭载多台传感器,在获取垂直影像同时能够获得包含地面物体的立面纹理信息的倾斜影像。该技术生成的高精度三维模型中,可以实现对高度、长度、面积等信息的直接量取。航空倾斜摄影测量技术相对于传统摄影测量技术手段优势巨大,但是其数据处理的难度相应提高了很多:倾斜影像因为大角度的倾斜角而导致影像之间存在地物的大仿射变形、严重透视畸变等问题;而分辨率变化大、影像大角度旋转、地物相互遮挡等问题也难以处理;倾斜摄影测量仪器同时搭载的多台传感器也产生了大量的数据冗余。这些客观情况的存在,给影像处理技术提出了很高要求,航空影像的处理算法需要重新设计。针对以上问题,本文研究了数字图像处理中的特征提取与匹配算法,尝试解决倾斜影像变形大导致的大失配问题。文章的研究工作和创新性主要体现在以下几个方面:一、针对倾斜影像的匹配这一主题,对倾斜摄影测量以及数字图像处理中特征提取以及匹配的基本知识进行介绍,重点分析了倾斜影像的特点以及当前流行的特征提取与匹配算法的局限性。二、重点研究了尺度不变特征算法(Scale-Invariant Feature Transform,SIFT)以及最大稳定极值区域算法(Maximally Stable Extremal Regions,MSER)之后,提出一种结合MSER和SIFT的SMSER算法:对MSER算法提取的仿射不变区域,经过拟合后用SIFT描述符进行描述,根据得到的SIFT描述符进行影像匹配。针对该方法,本文设置了相应的实验,验证了SMSER算法具有一定效果。
[Abstract]:Since the information revolution broke out, the storage, transmission and representation of information have changed qualitatively, which directly leads to the geographical information in data acquisition. Tilt photogrammetry is a great breakthrough in aerial photogrammetry technology, by carrying multiple sensors on the same flight platform. The vertical image can be obtained at the same time the tilt image containing the elevation texture information of the ground object can be obtained. In the high-precision 3D model generated by this technique, the height and length can be realized. Aerial tilt photogrammetry has a great advantage over traditional photogrammetry. However, the difficulty of data processing is much higher: the large affine deformation of ground objects and the serious perspective distortion exist in the inclined image because of the large angle of tilt; But the resolution change is big, the image big angle rotates, the ground object mutually blocks and so on the question also is difficult to deal with; Many sensors used in tilting photogrammetry instruments also produce a lot of data redundancy. The existence of these objective conditions puts forward high requirements for image processing technology. The algorithm of aeronautical image processing needs to be redesigned. In view of the above problems, this paper studies the feature extraction and matching algorithm in digital image processing. Try to solve the problem of large mismatch caused by large deformation of inclined image. The research work and innovation are mainly reflected in the following aspects: first, aiming at the subject of tilt image matching. The basic knowledge of feature extraction and matching in tilt photogrammetry and digital image processing is introduced. The characteristics of tilt image and the limitations of popular feature extraction and matching algorithms are analyzed. 2. The scale-Invariant Feature Transform, a scale-invariant feature algorithm, is studied emphatically. SIFT and the maximum Stable Extremal region algorithm (MSER). A SMSER algorithm combining MSER and SIFT is proposed. The affine invariant region extracted by MSER algorithm is described by SIFT descriptor after fitting. The image matching is carried out according to the obtained SIFT descriptors. In this paper, the corresponding experiments are carried out to verify the effectiveness of the SMSER algorithm.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P231

【相似文献】