无人机图像自动拼接问题研究

发布时间：2019-02-18 16:00

【摘要】：近年来,无人机低空遥感系统因其自身独特的优势,逐渐成为常规遥感系统(卫星遥感、航天航空遥感)的有效补充,作为一种新的对地观测手段广泛应用于各个领域,尤其是在应急灾害救援中发挥着不可替代的作用。在近期云南发生的盈江地震、贡山泥石流、洱源特大泥石流、彝良地震等应急救灾事件中,无人机低空遥感系统第一时间获取了分辨率为0.2m～0.4m的影像数据。但这些影像都存在像幅小、数量多、数据量大、畸变严重、重叠度不规则等问题,为了获得良好的视觉效果,方便灾害信息的有效提取和分析,必须对其进行快速的拼接,并且能与原有的地图资料通过google earth等软件进行叠加分析,实现快速的空间定位,及时掌握灾情信息,指导救援工作。为了在应急救灾中充分发挥无人机遥感影像的优势,本文做了以下方面的研究： (1)分析无人机遥感系统的误差来源；根据数码相机的特点及畸变模型,研究如何改正原始影像的畸变差；研究无人机的空中三角测量及流程； (2)通过深入研究无人机影像配准算法的原理、特点及流程,将基于特征点匹配的SIFT算法应用到无人机影像的自动匹配中,并分析基于坐标信息和基于SIFT特征这两种图像拼接思路的优缺点,融合这两种方法的优点,提出改进的无人机遥感影像拼接技术方法； (3)在进行影像融合时,分析影像融合的算法原理和优缺点,提出一种改进的加权融合算法,重新计算出归一化处理后参考影像和待拼接影像重叠区域的平均像素值,再分别处理影像重叠区域的两侧,使拼接全景图颜色实现自然过渡。 (4)采用C#语言,在Visual Studio201勺平台上成功利用改进的影像拼接技术方法对数十张影像进行拼接融合,获得的拼接全景图不仅具有坐标信息,还可以与google earth等软件进行叠加分析,进行快速的空间定位,指导救援工作,具有很高的实用价值。
[Abstract]:In recent years, unmanned aerial vehicle (UAV) low altitude remote sensing system has become an effective supplement to conventional remote sensing system (satellite remote sensing, aerospace remote sensing) because of its unique advantages. As a new means of earth observation, it has been widely used in various fields. Especially in the emergency disaster relief plays an irreplaceable role. In the recent Yingjiang earthquake, Gongshan debris flow, Eryuan mega-debris flow, Yiliang earthquake and other emergency disaster relief events, the UAV low-altitude remote sensing system obtained the image data with resolution of 0.2m~0.4m in the first time. However, there are many problems in these images, such as small image size, large amount of data, serious distortion, irregular overlap degree, etc. In order to obtain good visual effect and facilitate the effective extraction and analysis of disaster information, it must be quickly stitched together. And it can be superimposed with the original map data by google earth and other software to realize the rapid spatial positioning, grasp the disaster information in time, and guide the rescue work. In order to give full play to the advantages of UAV remote sensing image in emergency disaster relief, this paper has done the following research: (1) analyze the error source of UAV remote sensing system; According to the characteristics and distortion model of digital camera, this paper studies how to correct the distortion difference of the original image, and studies the aerial triangulation and flow chart of UAV. (2) by studying the principle, characteristics and flow of UAV image registration algorithm, SIFT algorithm based on feature point matching is applied to UAV image automatic matching. The advantages and disadvantages of the two image stitching methods based on coordinate information and SIFT feature are analyzed, and the advantages of these two methods are fused, and an improved remote sensing image mosaic method for UAV is proposed. (3) in the process of image fusion, the principle, advantages and disadvantages of the image fusion algorithm are analyzed, and an improved weighted fusion algorithm is proposed to recalculate the average pixel value of the overlapped region of the reference image and the image to be stitched after normalized processing. Then the two sides of the overlapped region are processed separately, so that the color of the stitched panoramic image can be transited naturally. (4) using C # language, the improved image stitching technology is successfully used in Visual Studio201 platform to fuse dozens of images. The mosaic panoramic images not only have coordinate information, It can also be superimposed with software such as google earth to locate space quickly and guide rescue work. It is of high practical value.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P231;P237

【参考文献】