Method for generating DSM (Digital Subscriber Model) through grid optimization by using satellite stereo image pair

Abstract

The invention relates to a method for generating a DSM (Digital Subscriber Model) through grid optimization by utilizing a satellite stereo image pair. The method comprises the following steps: firstly, generating a rough original three-dimensional grid according to a multi-scene linear array satellite stereo image pair, and taking the rough original three-dimensional grid as input; then representing a relation between an object point and an image point according to an RPC model, calculating a re-projection error, a Laplacian gradient and the like, respectively forming an energy function luminosity consistency energy term and a gradient smoothing energy term by taking the two calculated quantities as main indexes, and completing the construction of an energy function; continuously carrying out subdivision on the three-dimensional grid according to a self-adaptive threshold value; and a solution process is optimized by adopting a discrete-first optimization-second method. According to the method, the problem of size adaptability of densely matched windows in a traditional generation mode and the implicit problem of parallax consistency limitation in the windows are avoided, and the precision is higher.

Description

technical field [0001] The invention belongs to the field of three-dimensional modeling, and in particular relates to a method for generating DSM by grid optimization using satellite stereo pairs. Background technique [0002] In the production process of map products, DSM (Digital Surface Model) is the extraction of DEM (Digital Elevation Model), the generation of contour lines, the extraction and reconstruction of buildings, the production of real projection images and the update of geographic information. An important early data source for other work. The existing DSM generation methods mainly include laser scanning (Light Detection and Ranging) and photogrammetry. [0003] The method of directly acquiring DSM by laser scanning has high acquisition efficiency and is not limited by sunlight and weather, but the directly acquired 3D point cloud has some problems such as interference point removal, fracture line acquisition, and discontinuity-maintained building surface int...

AI Technical Summary

Problems solved by technology

[0003] The method of directly acquiring DSM by laser scanning has high acquisition efficiency and is not limited by sunshine and weather, but the 3D point cloud obtained directly has problems such as interference point removal, fault line acquisition, and building surface interpolation with discontinuity maintenance. Dealing with technical "bottlenecks"

[0004] The key technology that the photogrammetry method relies on to obtain DSM is the dense matching of images, which is generally carried out by using window correlation methods, but it faces the problem of adaptability to the size of the correlation window and the implicit limitation of disparity consistency within the window

The method of semi-global matching (SGM: semi-global matching) achieves an effect comparable to that of laser scanning, but its computational cost and memory overhead are relatively large, depending on the maximum parallax search range and the number of pixels in the image.

Images