CN-121982227-A - Three-dimensional image modeling method and device based on plane data

CN121982227ACN 121982227 ACN121982227 ACN 121982227ACN-121982227-A

Abstract

The invention provides a three-dimensional image modeling method and device based on plane data, which comprises the steps of carrying out full-color/multispectral wave band fusion, bit depth adjustment, thinning cloud, image enhancement, fringe removal and dodging on satellite mapping images to realize real color restoration of the images, carrying out splicing pretreatment on a plurality of small orthographic satellite mapping images after the real color restoration, carrying out image registration on the images to be registered and a reference image, carrying out image geometric correction on the images after the registration, carrying out image mosaic treatment on the images after the image geometric correction to obtain a single large scene image, carrying out three-dimensional image matching, building mask extraction, three-dimensional building modeling and three-dimensional building post-treatment on the single large scene image to obtain a three-dimensional digital surface model, and completing ground three-dimensional real image modeling. By applying the technical scheme of the invention, the technical problem that the existing two-dimensional mapping data are difficult to meet the application requirements of increasingly-growing three-dimensional scenes is solved.

Inventors

LI YU
ZHAO JINGCHAO
CHU XIANYING
CHEN CONG
XU BAOHUA

Assignees

北京空天技术研究所

Dates

Publication Date: 20260505
Application Date: 20251215

Claims (10)

1. The stereo image modeling method based on the plane data is characterized by comprising the following steps of: Step one, full color/multispectral wave band fusion, bit depth adjustment, thin cloud removal, image enhancement, stripe removal and dodging are carried out on a satellite mapping image, so that true color recovery of the satellite mapping image is realized; Performing stitching pretreatment on a plurality of small orthosatellite mapping images with the real colors recovered, performing image registration on the images to be registered after stitching pretreatment and a reference image, performing image geometric correction on the registered images, and performing image mosaic treatment on a plurality of images after image geometric correction to obtain a single large scene image; And thirdly, carrying out stereoscopic image matching, building mask extraction, three-dimensional building modeling and three-dimensional building post-processing on the single large scene image to obtain a three-dimensional digital surface model, and completing stereoscopic image modeling based on plane data.
2. The method of claim 1, wherein the fusion method for full-color/multispectral band fusion of satellite mapping images comprises IHS transformation, principal component transformation, weighted product, ratio transformation, wavelet transformation, high-pass filtering, BROVERY, PANSHARP fusion method combining GRB and IHS transformation.
3. The method for modeling a stereoscopic image based on planar data as defined in claim 2, wherein the bit depth adjustment of the satellite mapping image specifically includes processing the raw data of different bit depths into an 8bit image for display, the data information being required to be maximally retained according to the statistical distribution of the data.
4. The method of modeling a stereoscopic image based on planar data as defined in claim 3, wherein striping the satellite mapping image comprises introducing edge weighting factors on the constraint of the vertical direction of the striping, and solving and optimizing the proposed model by an Alternate Direction Multiplier Method (ADMM) to complete the striping of the satellite mapping image.
5. The stereo image modeling method based on planar data according to any one of claims 1 to 4, wherein in the second step, image registration is performed using a deep learning method including PointNet and OA-NET, and a supersue algorithm based on an attention mechanism and LofTR, coTR, clusterGNN based on a transducer.
6. The method according to claim 5, wherein in the second step, the geometric correction is performed by using RPC model (rational polynomial) correction, feature point fine correction, and the like.
7. The method of three-dimensional image modeling based on planar data according to claim 6, wherein in the second step, performing image mosaic processing on the plurality of geometrically corrected images to obtain a single large scene image specifically comprises: Splicing the small orthographic images subjected to the geometric correction according to the pixel and geographic position relationship to obtain an original large mosaic image; Downsampling the original large mosaic image according to pyramid level, calculating to obtain downsampling level under the condition of designating the size of the working image, and calculating downsampling thumbnail of the original large mosaic image; finishing mosaic line calculation, mask image calculation and color gradient smoothing on the downsampled thumbnail to generate a small-size mosaic preview image; performing color processing on the small-size mosaic preview image to obtain preview images with different styles and colors; and mapping the color from the preview image to the original large mosaic image by using Laplacian pyramid color fusion and multiband fusion technologies and adopting a blocking processing technology method to obtain a single large scene image.
8. The method for modeling a stereoscopic image based on planar data as claimed in claim 7, wherein said step three specifically comprises: based on the satellite image stereopair and SRTM elevation data, carrying out stereoscopic image matching on a single large scene image to generate an idealized earth surface model LOD0; building mask extraction is carried out on the basis of an idealized earth surface model LOD0, and building heights are added to building masks to generate a simple building model L0D1; On the basis of the simple building model L0D1, generating a refined building model L0D2 by combining an open street map (OpenStreetMap, OSM); And carrying out three-dimensional modeling post-processing on the building model L0D2 with the refinement, and generating a three-dimensional earth surface model.
9. Computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the steps of the computer program for implementing a method for three-dimensional image modeling based on planar data according to any of the claims 1 to 8.
10. A computer-readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the planar data based stereo image modeling method according to any of claims 1 to 8.

Description

Three-dimensional image modeling method and device based on plane data Technical Field The invention relates to the technical field of image processing, in particular to a three-dimensional image modeling method and device based on plane data. Background With the rapid development of aerospace, sensors and communication technologies, satellite remote sensing has made great progress, and has become an important means for earth observation and environmental monitoring. The geographical mapping technology has been developed rapidly by means of efficient and large-scale observation provided by satellite remote sensing. The satellite remote sensing can cover a wide geographical area in a short time, the mapping efficiency and range are greatly improved, the updating speed of satellite remote sensing data is high, latest earth surface information can be provided, the satellite remote sensing data is particularly important for rapidly changing areas, the optical resolution of satellite remote sensing images is improved increasingly, the satellite remote sensing data can be used for finer observation and inversion of ground scenes, and the satellite multispectral data is enriched increasingly, so that the satellite remote sensing data has the extremely important advantage of fine analysis on the ground. So far, the development of more than half century is carried out, a large amount of remote sensing earth observation data are collected in various countries around the world, a large amount of two-dimensional mapping results of categories such as digital elevation, orthographic images, digital line drawings and the like are produced, and a large amount of two-dimensional mapping data results and resources are accumulated. However, with development of technology and expansion of application requirements, two-dimensional mapping data are difficult to meet the increasing application requirements of three-dimensional scenes. The three-dimensional geographic data has more important application value in the scenes such as terrain analysis, flood simulation, city planning, emergency disaster relief, situation awareness, fine simulation and the like. Therefore, how to construct a realistic three-dimensional scene model based on two-dimensional mapping data such as remote sensing images, topographic maps and the like, and solve the problems of automation, high quality and high efficiency production, and become a popular and urgent application demand. Disclosure of Invention The invention provides a three-dimensional image modeling method and device based on plane data, which can solve the technical problem that the existing two-dimensional mapping data are difficult to meet the application requirements of increasingly three-dimensional scenes. The invention provides a three-dimensional image modeling method based on plane data, which comprises the steps of firstly, carrying out full-color/multispectral wave band fusion, bit depth adjustment, thinning cloud, image enhancement, fringe removal and dodging on satellite mapping images to realize real color recovery of the satellite mapping images, secondly, carrying out splicing pretreatment on a plurality of small orthographic satellite mapping images subjected to real color recovery, carrying out image registration on images to be registered after the splicing pretreatment and reference images, carrying out image geometric correction on the registered images, carrying out image mosaic treatment on a plurality of images subjected to image geometric correction to obtain single large scene images, and thirdly, carrying out three-dimensional image matching, building mask extraction, three-dimensional building modeling and three-dimensional building post-treatment on the single large scene images to obtain a three-dimensional digital surface model, thereby completing three-dimensional image modeling based on plane data. Further, fusion methods employed for full-color/multispectral band fusion of satellite mapping images include IHS transforms, principal component transforms, weighted products, ratio transforms, wavelet transforms, high-pass filtering, BROVERY, PANSHARP fusion methods combining GRB and IHS transforms. Further, the bit depth adjustment of the satellite mapping image specifically comprises the step of processing the original data with different bit depths into an 8-bit image which is convenient to display, and the data information is required to be maximally reserved according to the statistical distribution of the data. Further, striping the satellite survey image specifically includes introducing edge weight factors to the constraints of the vertical direction of the stripes, solving and optimizing the proposed model by an Alternating Direction Multiplier Method (ADMM) to complete the striping of the satellite survey image. Further, in step two, image registration is performed using a deep learning method, which includes PointNet and OA-NET, and a superstate algorith