基于多角度的多視影像真正射影像生成技術(shù)研究
發(fā)布時間：2019-02-17 19:45
【摘要】：數(shù)字正射影像(DOM)消除了地形起伏和相機姿態(tài)的影響,具有空間地物信息直觀、全面等特點,能夠作為地圖進行精確量測。數(shù)碼技術(shù)和傳感器技術(shù)的發(fā)展,促進了正射影像在城市規(guī)劃、GIS等領(lǐng)域的應(yīng)用。但是,數(shù)字正射影像仍舊存在許多問題,幾何方面的問題主要表現(xiàn)在人工建筑物傾斜,建筑物重影以及空間地物缺失等。由于DOM制作時并未考慮這些問題,因此正射影像糾正的準確性和完整性受到影響。如將全息矢量地圖數(shù)據(jù)與正射影像進行疊加,會出現(xiàn)建筑物被道路矢量線橫穿以及建筑物、道路等矢量輪廓與其自身產(chǎn)生錯位無法疊合等問題,影響其作為底圖的視覺分析功能。這些問題產(chǎn)生的關(guān)鍵原因是建筑物與空間地物之間存在遮擋關(guān)系。由于傳統(tǒng)正射影像應(yīng)用面臨的這些問題,真正射影像作為新的影像地圖產(chǎn)品應(yīng)運而生。真正射影像可以同時消除地形和重要空間地物目標投影的變形,保持了直視角度的地表景觀,影像色彩均勻,充分發(fā)揮了正射影像應(yīng)有的功能,解決了城市大比例尺正射影像陰影與高大建筑物對附近地表信息的遮擋等問題。由于真正射影像制作過程中最為關(guān)鍵的技術(shù)是遮蔽區(qū)域的檢測,不同的真正射影像制作技術(shù)的差異取決于遮蔽檢測方法,因此,大量學者將研究重點集中在遮擋檢測上。目前主要是利用數(shù)字建筑模型的方法和數(shù)字表面模型的方法進行遮擋檢測。本文在大量閱讀和深入研究真正射生成算法的基礎(chǔ)上,結(jié)合實驗數(shù)據(jù)源的特點,從兩個方向進行真正射影像生成技術(shù)研究:(1)基于數(shù)字城市三維模型的真正射影像生成技術(shù)首先,將獲取的多角度的多視影像進行數(shù)據(jù)預(yù)處理,快速準確獲取多視影像上的同名點坐標,進而空三加密得到地物的三維信息,實現(xiàn)規(guī)則與不規(guī)則建筑物幾何模型快速提取,完成數(shù)字城市三維模型重建。其次,針對正射影像幾何方面存在的問題,利用已獲取的建筑物矢量數(shù)據(jù),將遮擋區(qū)域的檢測及遮蔽區(qū)域的補償作為主要內(nèi)容進行真正射影像的幾何糾正,進而分析和研究真正射影像的制作,最后進行色彩一致性處理和影像的鑲嵌得到準確的真正射影像。(2)基于數(shù)字表面模型的真正射影像生成技術(shù)由于Z-buffer方法已被證明在檢測和處理多角度的多視影像的遮蔽區(qū)域是有效的,基于此,利用已有匹配獲得的數(shù)字表面模型進行遮蔽區(qū)域檢測,實現(xiàn)對遮蔽區(qū)域的完整標記,通過標記結(jié)果進行遮蔽區(qū)域的紋理補償工作,最后是色彩一致性處理以及影像的鑲嵌,至此,視覺上準確和美觀的真正射影像制作完成。對于真正射影像制作中遮蔽區(qū)域的信息補償,主要是利用相鄰航空影像上的可見區(qū)域,選擇最優(yōu)像素來補償遮蔽區(qū)域的信息缺失,在此基礎(chǔ)上進行后續(xù)的色彩一致性處理和影像的鑲嵌。
[Abstract]:Digital orthophoto image (DOM) eliminates the influence of terrain fluctuation and camera attitude. It has the characteristics of visual and comprehensive spatial features, and can be used as a map to measure accurately. The development of digital technology and sensor technology has promoted the application of orthophoto in urban planning, GIS and other fields. However, there are still many problems in digital orthophoto image, such as the inclination of artificial buildings, the shadow of buildings and the absence of space objects. The accuracy and completeness of orthophoto correction are affected by the fact that these problems are not taken into account in DOM production. If the holographic vector map data and orthophoto image are superimposed, there will be problems such as the building is crossed by the road vector line and the building, the contour of the vector such as the road cannot be overlapped with itself, and so on. It affects the function of visual analysis as a base map. The key cause of these problems is the occlusion relationship between buildings and space objects. Due to these problems in traditional orthophoto applications, real images as a new image map product came into being. The real image can eliminate the deformation of the terrain and the projection of the important space object simultaneously, maintain the surface landscape of the direct view angle, the image color is uniform, and the function of the orthophoto image is brought into full play. The problems such as the shadow of large scale orthophoto image and the occlusion of nearby surface information by tall buildings are solved. Because the most important technology in the process of real image production is the detection of shaded area, the difference of different real image making technology depends on the method of shadow detection. Therefore, a large number of scholars have focused their research on occlusion detection. At present, the methods of digital building model and digital surface model are mainly used to detect occlusion. On the basis of reading and deeply studying the real algorithm of projective generation, this paper combines the characteristics of experimental data sources. In this paper, we study the real image generation technology in two directions: (1) the real image generation technology based on the 3D model of digital city. Firstly, we preprocess the acquired multi-angle multi-view image. The coordinates of the same name points on the multi-view images are obtained quickly and accurately, and then the three-dimensional information of the objects is obtained by space-three encryption. The rules and the geometric models of irregular buildings are extracted quickly, and the reconstruction of the 3D models of digital cities is completed. Secondly, aiming at the problems existing in orthophoto geometry, using the obtained building vector data, the detection of occlusion area and the compensation of shaded area are taken as the main contents of geometric correction of real image. And then analyze and study the making of real images, Finally, color consistency processing and image mosaic are carried out to get accurate true image. (2) the real image generation technology based on digital surface model has been proved to be in the detection and processing of multiple angles because of the Z-buffer method. The masking area of the multi-view image is effective. Based on this, the existing matching digital surface model is used to detect the shaded area, the complete marking of the masking area is realized, and the texture compensation of the masking area is carried out by the marking result. Finally, color consistency processing and image mosaic, so that the visual accuracy and beauty of the real image production completed. For the information compensation of the shaded area in the real image making, it is mainly to use the visible area on the adjacent aerial image to choose the best pixel to compensate for the missing information in the shaded area. On this basis, the following color consistency processing and image mosaic are carried out.
【學位授予單位】：蘭州交通大學
【學位級別】：碩士
【學位授予年份】：2015
【分類號】：TP391.41
本文編號：2425530
[Abstract]:Digital orthophoto image (DOM) eliminates the influence of terrain fluctuation and camera attitude. It has the characteristics of visual and comprehensive spatial features, and can be used as a map to measure accurately. The development of digital technology and sensor technology has promoted the application of orthophoto in urban planning, GIS and other fields. However, there are still many problems in digital orthophoto image, such as the inclination of artificial buildings, the shadow of buildings and the absence of space objects. The accuracy and completeness of orthophoto correction are affected by the fact that these problems are not taken into account in DOM production. If the holographic vector map data and orthophoto image are superimposed, there will be problems such as the building is crossed by the road vector line and the building, the contour of the vector such as the road cannot be overlapped with itself, and so on. It affects the function of visual analysis as a base map. The key cause of these problems is the occlusion relationship between buildings and space objects. Due to these problems in traditional orthophoto applications, real images as a new image map product came into being. The real image can eliminate the deformation of the terrain and the projection of the important space object simultaneously, maintain the surface landscape of the direct view angle, the image color is uniform, and the function of the orthophoto image is brought into full play. The problems such as the shadow of large scale orthophoto image and the occlusion of nearby surface information by tall buildings are solved. Because the most important technology in the process of real image production is the detection of shaded area, the difference of different real image making technology depends on the method of shadow detection. Therefore, a large number of scholars have focused their research on occlusion detection. At present, the methods of digital building model and digital surface model are mainly used to detect occlusion. On the basis of reading and deeply studying the real algorithm of projective generation, this paper combines the characteristics of experimental data sources. In this paper, we study the real image generation technology in two directions: (1) the real image generation technology based on the 3D model of digital city. Firstly, we preprocess the acquired multi-angle multi-view image. The coordinates of the same name points on the multi-view images are obtained quickly and accurately, and then the three-dimensional information of the objects is obtained by space-three encryption. The rules and the geometric models of irregular buildings are extracted quickly, and the reconstruction of the 3D models of digital cities is completed. Secondly, aiming at the problems existing in orthophoto geometry, using the obtained building vector data, the detection of occlusion area and the compensation of shaded area are taken as the main contents of geometric correction of real image. And then analyze and study the making of real images, Finally, color consistency processing and image mosaic are carried out to get accurate true image. (2) the real image generation technology based on digital surface model has been proved to be in the detection and processing of multiple angles because of the Z-buffer method. The masking area of the multi-view image is effective. Based on this, the existing matching digital surface model is used to detect the shaded area, the complete marking of the masking area is realized, and the texture compensation of the masking area is carried out by the marking result. Finally, color consistency processing and image mosaic, so that the visual accuracy and beauty of the real image production completed. For the information compensation of the shaded area in the real image making, it is mainly to use the visible area on the adjacent aerial image to choose the best pixel to compensate for the missing information in the shaded area. On this basis, the following color consistency processing and image mosaic are carried out.
【學位授予單位】：蘭州交通大學
【學位級別】：碩士
【學位授予年份】：2015
【分類號】：TP391.41
【參考文獻】
相關(guān)博士學位論文 前1條
1 許彪;基于航空影像的真正射影像制作關(guān)鍵技術(shù)研究[D];武漢大學;2012年
，本文編號：2425530
本文鏈接：http://sikaile.net/wenyilunwen/huanjingshejilunwen/2425530.html
最近更新
教材專著
