无人机真正射影像制作关键技术研究

发布时间：2018-11-14 20:31

【摘要】：航空影像纹理丰富、色彩逼真、直观性强,便于更新,具有良好的现势性,可用于数字测图,作为数据库更新的底图。普通的正射影像未考虑地表建筑物,因而影像上建筑物会出现倾斜、遮挡,正射影像在镶嵌时容易出现接边困难的问题。随着成像技术的不断提高,获取的影像分辨率也越来越高,在高大建筑物密集区,建筑物的倾斜遮挡表现尤为明显。建筑物出现倾斜遮挡后,影像难以与其他图件套合,严重时会失去地理参考价值,不能用作数据库底图。真正射影像能够很好地解决上述问题。它与普通正射影像的区别在于它考虑了建筑物的投影差,能够将建筑物的屋顶纠正到其对应的地基位置,解决普通正射影像上建筑物的遮挡问题和影像接边难的问题。真正射影像同时兼顾了影像的视觉特征和地图的几何精度,探究真正射影像制作具有重要意义。真正射影像在数字微分纠正时使用的是数字表面模型DSM。正射影像的精度很大程度取决于数字表面模型的精度。数字表面模型既描述了地表面的形态又顾及了地面建筑物的高程信息,可看作是数字高程模型DEM和数字建筑物模型DBM结合,因此精确的DEM和DBM是构建DSM的前提。本文以成都理工大学无人机影像为数据源,以学校高大建筑物密集区域为例,对真正射影像制作的关键技术进行了研究。主要研究内容为:(1)通过影像匹配技术提取了实验区的点云数据,以得到地面点、获取精确DEM为目的,对DEM滤波进行了探究。针对自动点云滤波不能完全得到精确的地面点,采用人工精细化点云编辑。以理工东苑建筑物为例,总结了一套针对高大建筑物密集区域的DEM精细化编辑方案。(2)基于原始数字表面模型和精编后的DEM制作正射影像,分析了存在的建筑物边缘变形和倾斜问题。选择基于精编DEM的正射校正影像中建筑物投影差最小的位置用于镶嵌,并优化镶嵌线,一定程度上解决了建筑物倾斜、遮挡的问题。为彻底解决建筑物倾斜问题,构建了描述建筑物表面的数字模型——数字建筑物模型(DBM),采用“DEM+DBM”结合的形式重新构建了数字表面模型,该模型相较于原始数字表面模型能更加准确地描述建筑物和地表形态。(3)研究了基于DSM正射校正的“重影”问题,实现了影像正射校正与遮蔽区域检测的并行处理。针对真正射影像镶嵌线的复杂性,总结了一套基于纹理补偿的镶嵌线优化方案,最终实现了真正射影像的无缝拼接,很好地解决了普通正射影像中建筑物倾斜、互相遮挡的难题。
[Abstract]:Aerial images are rich in texture, realistic in color, highly intuitive, easy to update, and have good present situation. They can be used in digital mapping and as the base map of database updating. The normal orthophoto image does not consider the surface building, so the buildings on the image will be inclined and occluded, and the problem of edge joining is easy to occur when orthophoto image is inlaid. With the improvement of imaging technology, the resolution of the obtained images is higher and higher, especially the obliquity of the buildings in the dense areas of tall buildings. It is difficult for the image to fit with other maps when the building is obliquely shaded, which will lose the geographical reference value and can not be used as the base map of the database. The real image can solve the above problem well. The difference between it and the normal orthophoto image is that it considers the projection difference of the building, can correct the roof of the building to its corresponding foundation position, and solves the problem of the block of the building and the difficulty of connecting the image edge on the common orthophoto image. The real image takes into account the visual features of the image and the geometric accuracy of the map at the same time, so it is of great significance to explore the making of the real image. The digital surface model DSM. is used in digital differential correction of real image. The accuracy of orthophoto image largely depends on the accuracy of digital surface model. The digital surface model not only describes the shape of the ground surface but also takes into account the elevation information of the ground building. It can be regarded as the combination of the digital elevation model DEM and the digital building model DBM. Therefore, accurate DEM and DBM are the premise of constructing DSM. This paper takes UAV image of Chengdu University of Technology as data source and takes the dense area of university tall buildings as an example to study the key technology of real image making. The main research contents are as follows: (1) the point cloud data of the experimental area are extracted by image matching technique, and the DEM filtering is studied in order to get the ground point and obtain the accurate DEM. Because the automatic point cloud filter can not get the accurate ground point completely, the point cloud is edited by manual refinement. Taking the Dongyuan building of science and technology as an example, this paper summarizes a set of DEM fine editing schemes for dense areas of tall buildings. (2) based on the original digital surface model and finely edited DEM, orthophoto images are made. The existing problems of building edge deformation and inclination are analyzed. The position of minimum projection difference of buildings in orthographic correction image based on DEM is selected for mosaic and the inlay line is optimized to solve the problem of obliquity and occlusion of buildings to a certain extent. In order to solve the problem of building tilt thoroughly, a digital model describing the building surface is built. The digital building model (DBM), reconstructs the digital surface model in the form of "DEM DBM". Compared with the original digital surface model, this model can describe the building and surface shape more accurately. (3) the problem of "double shadow" based on DSM orthophoto correction is studied, and the parallel processing of orthophoto correction and shaded area detection is realized. In view of the complexity of real image mosaic line, this paper summarizes a set of mosaic line optimization scheme based on texture compensation, and finally realizes the seamless mosaic of real image, which solves the problem of building tilt in normal orthophoto image. The difficulty of blocking each othe
【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P231

【参考文献】