Five-lens oblique camera TDOM shielding area repairing method

Abstract

According to the invention, a five-lens oblique camera is adopted to make a TDOM, and a new idea and a new method for texture repair of a sheltered area are provided. The method comprises the following steps: 1, analyzing a sheltering phenomenon on an image and the influence of a flight height and a baseline on the sheltered area; 2, solving exterior orientation elements of the image by adopting INPHO software in combination with the POS value of the image; 4, manually measuring and extracting building corner data in combination with exterior orientation elements of the images; 4, performing texture repair on the relative occlusion area by adopting an inclined image correction algorithm: selecting a downward-looking image after occlusion detection and correction as a main image, selecting an inclined image as an auxiliary image, and selecting a corresponding visible pixel to perform texture repair on the relative occlusion area by utilizing a nearest site rule; and 5, carrying out texture repair on the absolutely shielded area which is not repaired by using model-driven image repair, solving the problems of image repair blurring and unclear building boundary, and finally obtaining a TDOM image with a good visual effect.

Description

technical field [0001] The present application relates to a method for repairing a TDOM occlusion area, in particular to a method for repairing a TDOM occlusion area with a five-lens tilt camera, and belongs to the technical field of TDOM occlusion repair. Background technique [0002] Digital orthophotos have both the geometric accuracy of maps and the basic features of images. They are not only used in various urban geographic information systems as background images to evaluate other map data, but also can update digital line map data in real time to improve data reliability. sex. The orthophotos corrected based on the digital ground model are not all orthophotos. Since the digital ground model does not include the height of the target objects (buildings, trees, etc.) on the ground, only terrain fluctuations are considered without consideration of buildings and trees and other tall objects, and the aerial images are taken by the center projection method, the tall objects...

AI Technical Summary

Problems solved by technology

[0016] First, digital orthophotos use digital elevation models to perform orthorectification on oriented images, without considering spatial objects such as artificial buildings, resulting in the central projection deformation of these objects still existing. TDOM uses DSM to correct the oriented images. The image is orthorectified, and tall objects such as buildings are corrected to the correct position. At the same time, it has texture information and geometric attributes of the map. Occlusion is inevitable in TDOM, which affects the basic functions of the image. In traditional TDOM production In the process, the occlusion area on the image is reduced by increasing the lateral and heading overlap between adjacent images, and the adjacent overlapping images are used to repair the texture of the occlusion area, while the absolute occlusion area in TDOM is not overdone. More attention and research, more directly using black pixels to fill, the production cost of the image increases with the increase of the overlap of adjacent images, increasing the overlap of adjacent images is not advisable in actual production, and in order to To improve the visual quality of the image, it is not feasible to simply fill the absolute occlusion area. The existing technology does not have a good solution to the occlusion, and lacks an efficient method for repairing the TDOM occlusion area;

[0017] Second, the occlusion detection methods in the prior art often cannot take into account the calculation efficiency and detection accuracy of the algorithm. On the one hand, the detection results of the method with high calculation efficiency are not accurate enough or the accuracy is not good, and a relatively large storage space is required; On the one hand, methods that have accurate detection results and do not require additional storage space consume a lot of detection time;

[0018] Third, the disadvantages of relative occlusion compensation based on adjacent orthophoto images in the existing technology include: First, the visibility of the occluded area of ​​the image is related to factors such as the height of the photographic center, the photographic baseline, and the height of buildings, and not all occluded areas Corresponding patched pixels can be found on other adjacent overlapping images, and some occluded areas are not visible on the image, forming an absolute occluded area. Unlike aerial images, which have adjacent overlapping images, there is no adjacent image available as an auxiliary image for adjacent image texture repair when performing texture repair in occluded areas; the third is to use adjacent images for relative occlusion texture The prerequisite for inpainting is to have a certain degree of overlap between adjacent images. If you want to make more occluded areas visible, you must also increase the overlap between images. The overlap between images does not meet the requirements for inpainting in occluded areas. It is not feasible to use the adjacent image repair method. In this case, the multi-view images in the combined camera can be used as auxiliary images to repair the main image;

[0019] Fourth, the first type of absolute occlusion compensation algorithm based on image repair in the prior art uses the principle of diffusion to repair, but the repaired area is often unclear and blurred; the second type maintains the integrity of image information and the image is not easy to Blurring occurs, but due to the limitations of the self-repair algorithm, the structural characteristics of buildings and other objects cannot be well maintained, and image breaks are prone to occur. Not very ideal, either the repair boundary is not clear, or the texture information is blurry and unrealistic

Moreover, it is even more unreliable in the face of large-area image defects. The algorithm that combines texture-based and structure-based images cannot achieve ideal repair results for the occluded areas of orthophotos. The boundaries of information defects are usually consistent with buildings. The edges are adjacent, and the edges of the building have relatively strong structural features. If the edge structure information is used as an important impression factor in the image restoration algorithm, it will be more wrong to fill the texture of the building roof into the shaded area.

Images