CN-121982570-A - County blue-green gray facility carbon pattern space fine recognition method based on machine learning

CN121982570ACN 121982570 ACN121982570 ACN 121982570ACN-121982570-A

Abstract

The invention discloses a county blue-green gray facility carbon pattern space fine recognition method based on machine learning, and relates to the technical field of carbon pattern recognition, comprising the steps of recognizing county satellite images, obtaining a plurality of grids, and carrying out color marking to obtain a blue region, a green region and a gray region, wherein the blue region is a water body region, the green region is a vegetation region and the gray region is an infrastructure region; the grid correction method comprises the steps of carrying out initial carbon balance marking on grids, correcting the initial carbon balance marking based on the functional attribute of the grids to obtain corrected carbon balance marking, calculating and obtaining corrected carbon quantification results of the grids, obtaining comprehensive confidence based on the synchronicity of the corrected carbon quantification results of adjacent grids and the synchronicity of the corrected carbon quantification results of grids with the same color mark, correcting the corrected carbon quantification results, and obtaining fine identification results of county blue-green gray facility carbon pattern space. The invention solves the technical problem of inaccurate identification of the carbon pattern in the prior art.

Inventors

WANG TIAN
ZHU XIAOQING
LIU CHENGMING
ZHAO QINFENG
DUAN ZIYU
ZHANG GAOSHAN

Assignees

杭州市拱墅区工大未来技术研究院

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (10)

1. The county blue-green gray facility carbon pattern space fine recognition method based on machine learning is characterized by comprising the following steps of: Identifying county satellite images, obtaining a plurality of grids, and performing color marking to obtain a blue area, a green area and a gray area, wherein the blue area is a water body area, the green area is a vegetation area and the gray area is an infrastructure area; Performing initial carbon balance marking on the grid, and correcting the initial carbon balance marking based on the functional attribute of the grid to obtain a corrected carbon balance marking; and calculating and acquiring a corrected carbon quantification result of the grid based on the corrected carbon balance marks, acquiring comprehensive confidence based on the synchronicity of the corrected carbon quantification result of the adjacent grid and the synchronicity of the corrected carbon quantification result of the grid with the same color mark, correcting the corrected carbon quantification result, and acquiring a carbon quantification result as a county blue-green gray facility carbon pattern space fine recognition result.
2. The machine learning-based county blue-green gray facility carbon pattern space fine recognition method of claim 1, wherein recognizing the county satellite image, obtaining a plurality of grids, and performing color marking to obtain a blue region, a green region and a gray region, comprises: acquiring a county satellite image; Gridding the county satellite image to obtain a plurality of grids, wherein the grid size is obtained based on county area; And based on machine learning, performing color marking on the obtained grid to obtain a blue region, a green region and a gray region, wherein the blue region is a water body region, the green region is a vegetation region and the gray region is an infrastructure region.
3. The machine learning based county blue-green gray facility carbon pattern space fine identification method of claim 1, wherein the initial carbon balance marking of the grid comprises: and marking the initial carbon balance of the grid based on the color mark of the grid, wherein the initial carbon balance mark comprises an initial carbon fixation rate and an initial carbon emission rate.
4. The machine learning based county blue-green gray facility carbon pattern space fine identification method of claim 1, wherein modifying the initial carbon balance mark based on the functional attribute of the grid, obtaining a modified carbon balance mark comprises: Acquiring functional attributes of the grid, wherein the functional attributes comprise the functional attributes of main facilities with the largest occupied area in the grid; and correcting the initial carbon balance mark based on the functional attribute of the grid to obtain a corrected carbon balance mark.
5. The machine learning-based county blue-green gray facility carbon pattern space fine recognition method according to claim 1, wherein the calculating and obtaining the corrected carbon quantization result of the grid based on the corrected carbon balance mark, and the obtaining the comprehensive confidence based on the corrected carbon quantization result synchronicity of the adjacent grid and the corrected carbon quantization result synchronicity of the same color mark grid, correcting the corrected carbon quantization result, and obtaining the carbon quantization result as the county blue-green gray facility carbon pattern space fine recognition result, comprises: Calculating and obtaining a corrected carbon quantification result of the grid based on the corrected carbon balance mark, wherein the corrected carbon quantification result comprises a corrected carbon emission rate or a corrected carbon fixation rate; Acquiring a proximity confidence based on the synchronicity of the corrected carbon quantification results of the grid and the adjacent grid; based on the color marks of the grids, obtaining grids of the same kind, and based on the synchronicity of the corrected carbon quantification results of the grids and the grids of the same kind, obtaining the confidence of the same kind; Based on the adjacent confidence coefficient and the similar confidence coefficient, acquiring comprehensive confidence coefficient, correcting the corrected carbon quantification result, and acquiring a carbon quantification result which is used as a county blue-green gray facility carbon pattern space fine recognition result.
6. The machine learning based county blue-green gray facility carbon pattern space fine recognition method of claim 5, wherein obtaining the proximity confidence parameter based on synchronicity of the revised carbon quantification results of the grid and the neighboring grid comprises: Acquiring 8 adjacent grids of the grids as adjacent grids; Based on machine learning, judging the synchronicity of the corrected carbon quantification results of the grids and the adjacent grids, and acquiring the adjacent confidence.
7. The machine learning based county blue-green gray facility carbon pattern space fine recognition method of claim 6, wherein determining the synchronicity of the revised carbon quantification results of the grid and the neighboring grid based on machine learning comprises: Acquiring a carbon quantification result set of the manual annotation grid, and evaluating the synchronicity of adjacent grids in the carbon quantification result of the manual annotation grid to acquire a synchronicity result set; Based on machine learning, constructing a synchronicity analyzer, taking a carbon quantification result set of the manual labeling grid as input, taking the synchronicity result as supervision, and training the synchronicity analyzer until convergence; and inputting the corrected carbon quantification results of the grids and the adjacent grids into the synchronicity analyzer to obtain synchronicity analysis results.
8. The machine learning based county blue-green gray facility carbon pattern spatial fine recognition method of claim 5, wherein obtaining a comprehensive confidence level based on said proximity confidence level and said like confidence level comprises; acquiring an adjacent confidence weight based on the same grid occupation ratio in adjacent grids of the grid as the grid color mark; subtracting the adjacent confidence coefficient weight from the calculated 1 to obtain the similar confidence coefficient weight; And weighting and calculating the adjacent confidence coefficient and the similar confidence coefficient based on the adjacent confidence coefficient weight and the similar confidence coefficient weight to obtain the comprehensive confidence coefficient.
9. The machine learning based county blue-green gray facility carbon pattern spatial fine recognition method of claim 1, further comprising: calculating the sum of the carbon quantification results of a plurality of grids to obtain county carbon quantification results; judging the deviation rate of the county carbon quantization result based on the historical county carbon quantization result; And judging the deviation direction based on the deviation rate, and acquiring a color mark correction parameter when the deviation rate is larger than a deviation rate threshold value, correcting the initial carbon balance mark and re-participating in the acquisition of the subsequent corrected carbon balance mark.
10. The machine learning based county blue-green gray facility carbon pattern space fine identification method of claim 9, wherein determining a deviation direction based on the deviation rate, when the deviation rate is greater than a deviation rate threshold, obtaining a color mark correction parameter comprises: And judging the deviation direction based on the deviation rate, and acquiring a color mark correction parameter when the deviation rate is larger than a deviation rate threshold value, wherein the gray correction parameter is larger than the green correction parameter and is larger than the blue correction parameter.

Description

County blue-green gray facility carbon pattern space fine recognition method based on machine learning Technical Field The invention relates to the technical field of carbon pattern recognition, in particular to a county blue-green gray facility carbon pattern space fine recognition method based on machine learning. Background In the global context of the carbon neutralization and sustainable development targets, county domains serve as key basic administrative and geographic units, and accurate accounting of carbon emission and carbon sink is an important basis for scientifically formulating emission reduction policies and implementing ecological management. Currently, carbon pattern evaluation for county scale mainly depends on a macroscopic evaluation method combining remote sensing image interpretation with statistical data. In the prior art, satellite images are generally utilized to perform large-scale vegetation index inversion or to perform coarse classification on land utilization/coverage types, so that empirical carbon emission or carbon sink coefficients are given to different types, and the regional carbon balance condition is estimated. However, such methods have obvious limitations in that it is difficult to finely identify the spatial pattern of the complex facility interweaving inside the county domain in terms of spatial resolution and classification accuracy, and erroneous judgment or omission is easily generated, resulting in inaccurate recognition of the carbon pattern. Disclosure of Invention The application provides a county blue-green gray facility carbon pattern space fine recognition method based on machine learning, which is used for solving the technical problem of inaccurate carbon pattern recognition in the prior art. In view of the above problems, the present application provides a county blue-green gray facility carbon pattern space fine recognition method based on machine learning, the method comprising: Identifying county satellite images, obtaining a plurality of grids, and performing color marking to obtain a blue area, a green area and a gray area, wherein the blue area is a water body area, the green area is a vegetation area and the gray area is an infrastructure area; Performing initial carbon balance marking on the grid, and correcting the initial carbon balance marking based on the functional attribute of the grid to obtain a corrected carbon balance marking; and calculating and acquiring a corrected carbon quantification result of the grid based on the corrected carbon balance marks, acquiring comprehensive confidence based on the synchronicity of the corrected carbon quantification result of the adjacent grid and the synchronicity of the corrected carbon quantification result of the grid with the same color mark, correcting the corrected carbon quantification result, and acquiring a carbon quantification result as a county blue-green gray facility carbon pattern space fine recognition result. One or more technical schemes provided by the application have at least the following technical effects or advantages: The application provides a county blue-green gray facility carbon pattern space fine recognition method based on machine learning, which is characterized in that the whole refinement degree of county scale carbon pattern space recognition, the accuracy of carbon balance estimation, the space consistency of the result and the system self-adaption capability are obviously improved by integrating the machine learning automatic meshing and color marking of satellite images, the refinement correction of initial carbon balance based on grid function attributes, the intelligent correction by utilizing the synchronization construction comprehensive confidence of the adjacent grids and similar grids to correct carbon quantization results, and the introduction of a feedback iterative optimization mechanism based on the whole deviation rate. Compared with the traditional method, the technical scheme provided by the application remarkably breaks through the problem of misjudgment or omission of the facility interweaving area in the traditional statistical estimation, and achieves the technical effect of realizing high-reliability refined identification of the blue, green and gray facility carbon patterns on the county scale. Drawings In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. Fig. 1 is a flow chart of a county blue-green gray facility carbon pattern space fine recognition method based on machine learning according to an embodiment of the application. Fig. 2 is a schematic flow chart