CN-122024073-A - Land utilization change information extraction method and system based on multi-source remote sensing image

CN122024073ACN 122024073 ACN122024073 ACN 122024073ACN-122024073-A

Abstract

The invention relates to the technical field of remote sensing and geographic information, and discloses a land utilization change information extraction method and system based on a multi-source remote sensing image. The method comprises the steps of obtaining optical or synthetic aperture radar remote sensing images with different time phases, carrying out geometric correction, radiation normalization, cloud removal restoration and format adaptation to generate a standardized image data set, selecting an adaptation algorithm according to image types, fusing spectrum and texture features to generate an initial change mask, and outputting SHP format change pattern spots and an accuracy report which accord with industry standards through pattern spot boundary optimization and base pattern registration verification. The system comprises a multi-source image input unit, a preprocessing unit, a change detection unit and a post-processing verification unit 4. The method and the device improve the precision, robustness and achievement normalization of land utilization change identification through full-flow automatic processing.

Inventors

LIU QI
ZHAO HUI
Liu Xiajing
LIU DANDAN
MA YANJUN
HU HONGJIANG

Assignees

河北省地质矿产勘查开发局第四水文工程地质大队(河北省地面沉降监测预警防治中心)
沧州渤海新区黄骅市自然资源和规划建设局

Dates

Publication Date: 20260512
Application Date: 20260409

Claims (10)

1. The land utilization change information extraction method based on the multi-source remote sensing image is characterized by comprising the following steps of: Acquiring remote sensing image data of at least two different time phases, wherein the remote sensing image data is in a TIFF format and comprises an optical remote sensing image or a synthetic aperture radar remote sensing image; Performing multisource remote sensing image preprocessing operation on the remote sensing image data, wherein the preprocessing operation comprises geometric correction, radiation normalization, cloud removal processing and format adaptation so as to generate a standardized remote sensing image data set which is aligned in space, consistent in radiation, free of cloud shielding and capable of retaining original space reference information; Performing land utilization change detection operation based on the standardized remote sensing image dataset, wherein the change detection operation comprises selecting an adaptive change detection algorithm according to an image type, extracting spectral features and texture features to construct a difference feature set, and generating a binarized initial change mask; And performing post-processing and verification operation of the change pattern spots on the initial change mask, wherein the post-processing and verification operation comprises pattern spot boundary optimization, base pattern spot registration verification and result output so as to generate a high-precision land utilization change pattern spot file and a change detection precision report which accord with industry standards.
2. The land use change information extraction method based on a multi-source remote sensing image according to claim 1, wherein performing a multi-source remote sensing image preprocessing operation on the remote sensing image data comprises: selecting ground control points to perform space coordinate matching on different simultaneous remote sensing images, unifying the images to the same geographic coordinate system by adopting a polynomial transformation model, keeping the space resolution of the corrected images consistent with that of the original images, and enabling the geometric position error to be smaller than the pixel; Respectively carrying out histogram matching treatment on each wave band of different time phase remote sensing images to enable the radiation distribution of the target image to be consistent with that of the reference image, or adopting a linear regression model to establish a radiation brightness mapping relation of corresponding wave bands between two phases of images, and carrying out radiation scale adjustment on the target image according to the radiation brightness mapping relation; identifying cloud and cloud shadow areas in an image based on a multi-threshold segmentation method, wherein the multi-threshold comprises a near infrared band reflectivity threshold, a blue light band reflectivity threshold and a normalized vegetation index threshold, and performing pixel restoration on the identified cloud areas by adopting a space-time interpolation algorithm, wherein the space-time interpolation algorithm performs weighted interpolation by utilizing pixel values of cloud-free areas in adjacent time phases at the same geographic position; And outputting the preprocessed image in a standardized TIFF format with original spatial reference information reserved, wherein the spatial resolution, projection parameters and a coordinate system are consistent with those of the reference image in the input image.
3. The land use change information extraction method based on multi-source remote sensing images according to claim 2, wherein performing a land use change detection operation based on the standardized remote sensing image dataset comprises: If the input image contains the synthetic aperture radar remote sensing image, adopting a deep learning model to carry out change detection; The image difference method comprises the steps of calculating a normalized vegetation index difference value, a brightness difference value or a water index difference value, performing principal component transformation on two-period images, extracting a difference value of a first principal component as a change judgment basis, and outputting a probability value of each pixel belonging to a change category by using a deep learning model which comprises a U-Net network structure or a CHANGENET network structure and is input as a splicing tensor of the two-period normalized remote sensing images; And extracting spectral features and texture features to construct a difference feature set, wherein the spectral features comprise gray values of all wave bands, normalized vegetation indexes, normalized water indexes and soil-adjusted vegetation indexes, and the texture features comprise contrast, correlation, energy and homogeneity calculated based on gray co-occurrence matrixes, and variance, entropy and gradient amplitude calculated based on local windows.
4. A land use change information extraction method based on multi-source remote sensing images as claimed in claim 3, wherein generating a binary initial change mask comprises: And (3) setting a fixed threshold value or a self-adaptive threshold value for the difference feature set, performing binarization processing, marking pixels with difference values larger than the threshold value as a change region, marking other pixels as non-change regions, or directly outputting a binarization mask by a deep learning model, wherein the pixel value of the change region is 1, and the pixel value of the non-change region is 0.
5. The land use change information extraction method based on multi-source remote sensing images according to claim 4, wherein the map spot boundary optimization comprises: And performing morphological closing operation on the initial change mask to fill the internal holes, performing connectivity analysis on edge pixels by adopting a region growing algorithm, determining an optimal boundary by combining self-adaptive threshold segmentation, eliminating isolated pixels and jagged edges, and ensuring continuous and complete image spot boundaries.
6. The land use change information extraction method based on multi-source remote sensing images according to claim 5, wherein the base map patch registration verification comprises: importing land use current state base map spot data in SHP format, performing space superposition analysis on the optimized change map spot and the base map spot, calculating the boundary coincidence degree of the two, and replacing the corresponding boundary of the change map spot with the boundary of the base map spot when the coincidence degree is more than 85%, thereby completing the map spot standardization processing.
7. The land use change information extraction method based on multi-source remote sensing images according to claim 6, wherein said outputting of results comprises: Generating a land utilization change pattern file in SHP format, wherein an attribute table of the land utilization change pattern file comprises a change type, a change area, a change start-stop time and a change confidence coefficient field, and simultaneously generating a change detection precision report, wherein the precision report comprises a omission factor, a false detection rate, overall precision and Kappa coefficient.
8. The land use change information extraction method based on the multi-source remote sensing image according to claim 1, wherein in the geometric correction, ground control points are uniformly distributed in the whole image domain, the polynomial transformation order is second order or third order, and parameter solving is performed by taking the minimum residual square sum as an objective function.
9. The land use change information extraction method based on multi-source remote sensing images according to claim 2, wherein the stable and unchanged area in the radiation normalization comprises bare soil or water body for fitting a linear regression model , For the value of the band of the target image, For the reference image band value, Is the slope of the slope, Is the intercept.
10. Land utilization change information extraction system based on multisource remote sensing image, characterized by comprising: the multi-source remote sensing image input unit is used for receiving at least two different-phase TIFF format remote sensing image data, wherein the remote sensing image data comprises an optical remote sensing image or a synthetic aperture radar remote sensing image; the multi-source remote sensing image preprocessing unit is used for performing geometric correction, radiation normalization, cloud removal processing and format adaptation on the remote sensing image data to generate a standardized remote sensing image data set; The land utilization change detection unit is used for selecting an adaptive change detection algorithm according to the image type based on the standardized remote sensing image data set, extracting spectral features and texture features and generating an initial change mask; And the change pattern spot post-processing and verifying unit is used for performing pattern spot boundary optimization, base pattern spot registration verification and result output on the initial change mask to generate a high-precision land utilization change pattern spot file and a change detection precision report.

Description

Land utilization change information extraction method and system based on multi-source remote sensing image Technical Field The invention belongs to the technical field of remote sensing and geographic information, and particularly relates to a land utilization change information extraction method and system based on a multi-source remote sensing image. Background With the rapid development of remote sensing technology and the wide acquisition of multi-source satellite data, land utilization change monitoring has become a key supporting means in natural resource management, ecological environment assessment and urban planning decisions. The remote sensing image is widely applied to the dynamic analysis of the land coverage with large scale and long time sequence by virtue of the advantages of wide coverage, short updating period, rich information dimension and the like. However, the land utilization change information extraction involves a complex image processing flow, and needs to integrate various sensor data such as optics, synthetic Aperture Radar (SAR) and the like, and perform accurate comparison at different time intervals, which puts high requirements on data preprocessing, change detection algorithm and result post-processing. A land utilization change information extraction method based on multi-source remote sensing images aims at identifying an area with changed earth surface coverage type by comparing two or more time-phase remote sensing images and generating a change pattern spot with definite boundary and attribute information. The core of the technical direction is how to realize high-consistency and high-robustness change region identification and refined expression under the complex interference conditions of atmospheric interference, illumination difference, geometric deviation, cloud cover and the like. The prior art still has the following problems that firstly, radiation and geometric characteristics of a multi-source remote sensing image are inconsistent due to sensor type, imaging angle and time difference, the coupling influence of atmospheric scattering, illumination change and geometric distortion is difficult to synchronously eliminate by a conventional preprocessing method, so that systematic deviation is introduced in subsequent change detection, secondly, a single change detection algorithm is difficult to adapt to heterogeneous characteristics of optical and SAR images, a deep learning model has potential, a large number of false detection or omission detection is generated under a complex ground object scene due to insufficient training samples or limited generalization capability, and thirdly, boundary saw teeth, holes and isolated noise pixels are commonly existed in an initial change mask, an effective post-processing mechanism and authoritative base image data linkage verification are lacked, so that final image spots are not in line with industry drawing specifications and are difficult to be directly used for a business GIS platform. The problems are particularly prominent in a large-scale and high-frequency land change investigation task, the accuracy, the efficiency and the practicability of remote sensing change detection results are severely restricted, and a full-flow solution integrating standardized pretreatment, self-adaptive detection and intelligent post-verification is needed. Disclosure of Invention The invention provides a land utilization change information extraction method and a land utilization change information extraction system based on a multisource remote sensing image, which are characterized in that by constructing an end-to-end remote sensing image processing and analyzing flow, the method integrates key technical links such as geometric correction, radiation normalization, cloud removal restoration, multi-mode change detection, pattern spot boundary optimization, base map registration verification and the like, and forms a standardized, automatic and high-precision land utilization change information extraction system. The method and the system can effectively eliminate non-uniform interference of the multi-source remote sensing image in the space-time dimension, accurately identify the land utilization change area, generate vector image spot achievements meeting the industry specification and remarkably improve the reliability and practicality of change detection. The invention provides a land utilization change information extraction method based on a multi-source remote sensing image, which comprises the following steps: Acquiring remote sensing image data of at least two different time phases, wherein the remote sensing image data is in a TIFF format and comprises an optical remote sensing image or a synthetic aperture radar remote sensing image; Performing multisource remote sensing image preprocessing operation on the remote sensing image data, wherein the preprocessing operation comprises geometric correction, radiation normalization,