CN-121980257-A - Photovoltaic remote sensing reference data set construction method and related device

CN121980257ACN 121980257 ACN121980257 ACN 121980257ACN-121980257-A

Abstract

The disclosure provides a photovoltaic remote sensing reference data set construction method and a related device, comprising the following steps: and acquiring a photovoltaic visible light image and a labeling data set thereof, constructing photovoltaic multi-mode data, and adapting a multi-mode remote sensing large model to obtain an adapted model. And analyzing the first multisource geospatial data set by using the model, generating a first grid annotation graph, constructing a photovoltaic annotation data set according to the first grid annotation graph, and further optimizing the model. And finally, inputting the second multisource geospatial data set into the optimized model, generating a second grid label graph and constructing a multi-mode photovoltaic data set. The method and the device can overcome the problem of false marks of the visible light model, and remarkably improve the accuracy of photovoltaic identification and the reliability of a data set.

Inventors

CUI MINGTAO
ZHAO WENBO
WANG CONGCONG
LI WANZHEN
SHI SHUAI

Assignees

国网信息通信产业集团有限公司

Dates

Publication Date: 20260505
Application Date: 20251203

Claims (10)

1. The construction method of the photovoltaic remote sensing reference data set is characterized by comprising the following steps of: Acquiring a photovoltaic visible light image of a multi-source data set and a labeling data set corresponding to the photovoltaic visible light image, and acquiring a photovoltaic multi-mode data set based on the photovoltaic visible light image and the labeling data set; adapting a preset multi-mode remote sensing large model based on the photovoltaic multi-mode data set to obtain an adapted multi-mode remote sensing model; Acquiring a first multisource geospatial data set, and analyzing the first multisource geospatial data set based on the adapted multi-mode remote sensing model to obtain a first grid label graph; constructing based on the first grid labeling graph to obtain a photovoltaic labeling data set; adapting the adapted multi-modal remote sensing model based on the photovoltaic labeling data set to obtain an optimized multi-modal remote sensing model; acquiring a second multisource geospatial data set, inputting the second multisource geospatial data set into the optimized multi-modal remote sensing model, and obtaining a second grid label graph output by the optimized multi-modal remote sensing model; and constructing based on the second grid label graph to obtain a multi-mode photovoltaic data set.
2. The method of claim 1, wherein the obtaining a photovoltaic multi-modal dataset based on the photovoltaic visible light image and the annotation dataset comprises: Sampling and screening the photovoltaic visible light image and the labeling data set to obtain a multi-source photovoltaic labeling set; and detecting the multi-source photovoltaic labeling set to obtain the photovoltaic multi-mode data set.
3. The method of claim 2, wherein detecting the multi-source set of photovoltaic labels to obtain the set of photovoltaic multi-modality data comprises: Detecting the multi-source photovoltaic labeling set, responding to the fact that the multi-source photovoltaic labeling set contains geographic coordinates, and correlating the multi-source photovoltaic labeling set containing the geographic coordinates to obtain multi-mode image data; And carrying out geographic registration and cutting on the multi-mode image data to obtain the photovoltaic multi-mode data set.
4. The method of claim 2, wherein detecting the multi-source set of photovoltaic labels results in the photovoltaic multi-modality dataset, further comprising: detecting the multi-source photovoltaic labeling set, and performing image division on the multi-source photovoltaic labeling set to obtain a standardized image unit in response to the multi-source photovoltaic labeling set not containing geographic coordinates; And fusing the standardized image units to obtain the photovoltaic multi-mode data set.
5. The method of claim 1, wherein analyzing the first multisource geospatial dataset based on the adapted multimodal remote sensing model results in a first grid label graph comprising: Performing geographic registration on satellite images of the multisource geospatial dataset to obtain registered image slices; labeling the registered image slices to obtain a multi-category labeling grid chart; Screening the multi-category labeling grid patterns to obtain a multi-mode photovoltaic structure data set; Extracting the multi-mode photovoltaic structure data set to obtain a building roof photovoltaic multi-mode data set; Screening the building roof photovoltaic multi-mode data set to obtain a roof photovoltaic structure data set; And the adapted multi-mode remote sensing model marks and adjusts the building roof photovoltaic multi-mode data set based on the roof photovoltaic structure data set to obtain the first grid mark graph.
6. The method of claim 1, wherein constructing based on the second grid-labeled graph results in a multi-modal photovoltaic dataset, comprising: Vectorizing the second grid label graph to obtain polygonal vector data; Screening based on the second grid annotation graph and the polygonal vector data to obtain in-doubt annotation image data and high-confidence annotation image data; correcting the suspicious marked image data to obtain corrected marked image data; And constructing based on the high-confidence labeling image data and the corrected labeling image data to obtain the multi-mode photovoltaic data set.
7. A photovoltaic remote sensing reference dataset construction apparatus, comprising: the multi-mode data set determining module is configured to acquire a photovoltaic visible light image of a multi-source data set and a labeling data set corresponding to the photovoltaic visible light image, and obtain a photovoltaic multi-mode data set based on the photovoltaic visible light image and the labeling data set; the adapted model determining module is configured to adapt a preset multi-mode remote sensing large model based on the photovoltaic multi-mode data set to obtain an adapted multi-mode remote sensing model; The first annotation map determining module is configured to acquire a first multi-source geospatial data set, analyze the first multi-source geospatial data set based on the adapted multi-mode remote sensing model, and acquire a first grid annotation map; The annotation data set determining module is configured to construct based on the first grid annotation graph to obtain a photovoltaic annotation data set; the optimized model determining module is configured to adapt the adapted multi-mode remote sensing model based on the photovoltaic labeling data set to obtain an optimized multi-mode remote sensing model; the second annotation map determining module is configured to acquire a second multi-source geospatial data set, input the second multi-source geospatial data set into the optimized multi-mode remote sensing model, and obtain a second grid annotation map output by the optimized multi-mode remote sensing model; and the photovoltaic data set determining module is configured to construct based on the second grid label graph to obtain a multi-mode photovoltaic data set.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of any one of claims 1 to 6 when the program is executed.
9. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1 to 6.
10. A computer program product comprising computer program instructions which, when run on a computer, cause the computer to perform the method of any one of claims 1 to 6.

Description

Photovoltaic remote sensing reference data set construction method and related device Technical Field The disclosure relates to the technical field of electric digital data processing, in particular to a photovoltaic remote sensing reference data set construction method and a related device. Background This section is intended to provide a background or context to the embodiments of the disclosure recited in the claims. The description herein is not admitted to be prior art by inclusion in this section. Along with the accelerated development of the novel power system, a large amount of high-proportion photovoltaic energy sources are connected into the novel power system. The photovoltaic energy source has the characteristics of intermittence, weather influence and the like, so that the photovoltaic energy source has a large influence on the stability of a power grid, and the photovoltaic remote sensing is a technology for automatically identifying, positioning and analyzing the spatial distribution, area and state of a ground photovoltaic power station and a photovoltaic panel by scanning a large-range area from the air through a sensor carried by a satellite or an aviation platform and combining a computer vision and artificial intelligence technology. However, in the related art, the data set construction technology only depends on visible light images, has single labeling category, is easily affected by interferents to cause false labeling, and lacks a multi-mode data fusion and automatic verification mechanism. Disclosure of Invention Accordingly, an objective of the present disclosure is to provide a method and related device for constructing a photovoltaic remote sensing reference data set, which at least partially solves one of the technical problems in the related art. In view of the above object, a first aspect of exemplary embodiments of the present disclosure provides a photovoltaic remote sensing reference dataset construction method, the method including: Acquiring a photovoltaic visible light image of a multi-source data set and a labeling data set corresponding to the photovoltaic visible light image, and acquiring a photovoltaic multi-mode data set based on the photovoltaic visible light image and the labeling data set; adapting a preset multi-mode remote sensing large model based on the photovoltaic multi-mode data set to obtain an adapted multi-mode remote sensing model; Acquiring a first multisource geospatial data set, and analyzing the first multisource geospatial data set based on the adapted multi-mode remote sensing model to obtain a first grid label graph; constructing based on the first grid labeling graph to obtain a photovoltaic labeling data set; adapting the adapted multi-modal remote sensing model based on the photovoltaic labeling data set to obtain an optimized multi-modal remote sensing model; acquiring a second multisource geospatial data set, inputting the second multisource geospatial data set into the optimized multi-modal remote sensing model, and obtaining a second grid label graph output by the optimized multi-modal remote sensing model; and constructing based on the second grid label graph to obtain a multi-mode photovoltaic data set. Based on the same inventive concept, a second aspect of exemplary embodiments of the present disclosure provides a photovoltaic remote sensing reference data set construction apparatus, including: the multi-mode data set determining module is configured to acquire a photovoltaic visible light image of a multi-source data set and a labeling data set corresponding to the photovoltaic visible light image, and obtain a photovoltaic multi-mode data set based on the photovoltaic visible light image and the labeling data set; the adapted model determining module is configured to adapt a preset multi-mode remote sensing large model based on the photovoltaic multi-mode data set to obtain an adapted multi-mode remote sensing model; the first grid annotation map determining module is configured to acquire a first multi-source geospatial data set, analyze the first multi-source geospatial data set based on the adapted multi-mode remote sensing model, and acquire a first grid annotation map; The annotation data set determining module is configured to construct based on the first grid annotation graph to obtain a photovoltaic annotation data set; the optimized model determining module is configured to adapt the adapted multi-mode remote sensing model based on the photovoltaic labeling data set to obtain an optimized multi-mode remote sensing model; The second grid annotation map determining module is configured to acquire a second multi-source geospatial data set, input the second multi-source geospatial data set into the optimized multi-mode remote sensing model, and obtain a second grid annotation map output by the optimized multi-mode remote sensing model; and the photovoltaic data set determining module is configured to construct based on the se