A high-resolution remote sensing image full-automatic atmospheric terrain radiation correction method and device

Abstract

The application discloses a kind of high remote sensing image full-automatic atmospheric terrain radiation correction method and device, belong to remote sensing science and technology field.The method includes: constructing full-automatic batch processing framework;Based on 6S radiation transfer model, develop atmospheric correction module, automatically retrieve atmospheric parameters and eliminate its influence;Based on digital elevation model (DEM) and VEGA model, develop terrain radiation correction module, compensate the radiation distortion caused by complex terrain;Key innovation lies in, construct joint correction framework and integrate topographic feature recognition algorithm, can automatically distinguish and select to execute single atmospheric correction or atmospheric-terrain joint correction process according to image terrain undulation degree.The application realizes full-chain automation batch processing from data input to result output, without manual intervention, significantly improve the processing efficiency of high-resolution image radiation correction, intelligent level and adaptability to different geomorphological scenes.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing science and technology, specifically relating to a fully automatic atmospheric topographic radiometric correction method and device for high-resolution remote sensing images. Background Technology

[0002] In the field of remote sensing science, remote sensing images are an important carrier for acquiring surface information. Atmospheric correction and topographic radiometric correction are two key steps in the image preprocessing process. Their purpose is to eliminate non-surface interference during the imaging process and restore the true spectral reflectance characteristics of ground objects.

[0003] Atmospheric correction aims to remove the effects of scattering, absorption, and path radiation of electromagnetic radiation signals by atmospheric molecules and aerosols. Without effective correction, the radiometric values ​​and spectral characteristics of images will be severely distorted, limiting subsequent quantitative analysis applications. Topographic radiometric correction focuses on eliminating differences in solar incident radiation caused by topographic factors such as slope and aspect. These differences are particularly significant in complex terrains such as mountains and hills, leading to huge variations in the brightness of the same type of land feature depending on its orientation, greatly interfering with the accuracy of feature identification and classification.

[0004] Currently, both academia and industry have conducted extensive research on these two correction techniques. Physical model-based correction methods (such as the Python library version of the 6S radiative transfer model and MODTRAN) have become mainstream due to their high accuracy and well-defined mechanisms. In topographic radiometric correction, numerous correction models based on digital elevation models have also been proposed. However, existing technologies still face several limitations in practical engineering applications: First, fragmented processes and low automation. Atmospheric correction and topographic radiometric correction are often treated as independent steps, requiring manual sequential execution by professionals. This involves extensive parameter setting and data preparation, resulting in low processing efficiency and difficulty in handling the batch processing needs of large-scale, multi-scene imagery. Second, a lack of scene adaptability. Most tools or systems employ fixed processing workflows and do not intelligently select the optimal correction strategy based on the image's own topographical characteristics (such as plains or mountains). For example, forcibly applying complex topographic radiometric correction to flat areas is not only ineffective but may also introduce errors. Third, high dependence on operator expertise. The accuracy of the entire processing chain heavily relies on the user's experience and judgment, making it prone to inconsistent results due to human error, thus presenting a high technical barrier.

[0005] Therefore, designing a high-resolution image atmospheric and topographic joint correction technology that can intelligently identify scenes, adaptively select correction processes, and achieve fully automated, batch operation has become a key requirement for improving the efficiency and intelligence level of remote sensing image preprocessing and promoting the large-scale operational application of remote sensing technology. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a fully automated atmospheric topographic radiometric correction method and apparatus for high-resolution remote sensing images. This method achieves full automation and intelligence in the remote sensing image correction process, improves correction efficiency and accuracy, enhances adaptability to different data sources and application scenarios, and reduces reliance on the professional skills of operators.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A fully automated atmospheric topographic radiometric correction method for high-resolution remote sensing imagery, the method comprising:

[0009] Step 1: Input high-resolution remote sensing image data, and automatically obtain the corresponding digital elevation model (DEM) data and atmospheric parameters;

[0010] Step 2: Perform automatic atmospheric correction on the high-resolution remote sensing image data based on the 6S radiative transfer model to obtain atmospherically corrected data;

[0011] Step 3: Automatically calculate terrain feature parameters based on the digital elevation model (DEM) data, and analyze the terrain complexity through a pre-trained discrimination model to output the discrimination probability. The pre-trained discrimination model is a two-branch neural network that includes a visual branch and a statistical branch.

[0012] Step 4: If the discrimination probability value is less than the first threshold, output the atmospherically corrected data; if the discrimination probability value is greater than or equal to the first threshold, call the VECA model to perform topographic radiation correction on the atmospherically corrected data and output the final surface reflectance product.

[0013] Furthermore, in step 1, based on the metadata information of the high-resolution remote sensing image data, the digital elevation model (DEM) data covering the image area is automatically retrieved and obtained from a preset external data source; and based on the imaging time and geographical location, the atmospheric parameters required for 6S model calculation are automatically extracted or inverted from the meteorological database or climate model.

[0014] Furthermore, the automatic atmospheric correction process in step 2 also includes: automatically detecting cloud, snow and thick haze coverage areas in the image before or during the 6S model calculation, and marking or masking them to avoid abnormal pixels from negatively affecting atmospheric parameter inversion and overall correction accuracy.

[0015] Furthermore, in step 3, the terrain feature parameters include: the average slope of the image coverage area calculated based on DEM data, and the coefficient of variation of the cosine value of the incident angle calculated based on the solar-terrain geometric relationship.

[0016] Furthermore, in the pre-trained discriminant model, the input of the visual branch is a combined matrix image composed of the slope map and illumination map generated based on the DEM, which is used to extract spatial texture features; the input of the statistical branch is a feature vector composed of the average slope, the coefficient of variation, and the solar zenith angle; after fusing the features of the visual branch and the statistical branch, a continuous probability value representing the necessity of terrain radiation correction is output by the classifier.

[0017] Furthermore, in step 4, a first threshold and a second threshold are set, and a hierarchical decision is made based on the discrimination probability value:

[0018] If the discrimination probability value is less than the first threshold, it is determined to be flat terrain, and terrain radiation correction is not performed; instead, atmospheric correction data is output.

[0019] If the discrimination probability value is greater than or equal to the first threshold and less than the second threshold, it is determined to be a slightly undulating terrain. The VECA model is called to perform terrain radiation correction on the atmospheric correction data to obtain and output the final correction result.

[0020] If the discrimination probability value is greater than or equal to the second threshold, it is determined to be complex terrain. The system automatically triggers the joint correction process, inputs the atmospheric correction data and the digital elevation model (DEM) data into the VECA model to perform terrain radiation correction, and obtains and outputs the final surface reflectance product.

[0021] Furthermore, the first threshold is 0.1, and the second threshold is 0.5.

[0022] On the other hand, the present invention provides a fully automatic atmospheric topographic radiometric correction device for high-resolution remote sensing images, comprising:

[0023] The input module is used to input high-resolution remote sensing image data and automatically obtain the corresponding digital elevation model (DEM) data and atmospheric parameters.

[0024] The atmospheric correction module is used to perform automatic atmospheric correction on the high-resolution remote sensing image data based on the 6S radiative transfer model to obtain atmospherically corrected data.

[0025] The discrimination module is used to automatically calculate terrain feature parameters based on the digital elevation model (DEM) data, analyze the terrain complexity through a pre-trained discrimination model, and output the discrimination probability. The pre-trained discrimination model is a two-branch neural network that includes a visual branch and a statistical branch.

[0026] The terrain correction module is used to output the atmospherically corrected data when the discrimination probability value is less than a first threshold; if the discrimination probability value is greater than or equal to the first threshold, the VECA model is called to perform terrain radiation correction on the atmospherically corrected data and output the final surface reflectance product.

[0027] Thirdly, the present invention provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned fully automatic atmospheric topographic radiometric correction method for high-resolution remote sensing images.

[0028] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned method for fully automatic atmospheric and topographic radiometric correction of high-resolution remote sensing images.

[0029] The beneficial effects of this invention are as follows:

[0030] Significantly improved processing efficiency and automation: By constructing a fully automated batch processing framework, the system can automatically complete the entire process from data import, parameter inversion, workflow selection to result output based on image metadata and terrain features, without manual intervention. This drastically reduces the processing time for a single scene image from several hours in traditional manual processing to minutes, making it particularly suitable for batch processing of large-scale remote sensing images and greatly improving work efficiency.

[0031] Enhanced Correction Accuracy and Scene Adaptability: This method innovatively integrates atmospheric correction and topographic radiometric correction, and intelligently selects single or combined correction processes based on the terrain recognition results. For complex terrains such as mountainous areas, an optimized topographic radiometric correction algorithm (VECA model, Vegetation-Terrain Coupled Radiometric Correction Model) is adopted, which can more accurately eliminate radiometric distortion caused by terrain undulations, making the radiometric values ​​of different slope aspects tend to be consistent, and significantly improving the spatial consistency and realism of the correction results under complex terrains.

[0032] Lowering the operational threshold and improving result stability: The system achieves automatic adaptation of the correction process through an algorithm framework, reducing reliance on the professional experience of operators and avoiding errors caused by improper manual parameter settings. This ensures the stability and reliability of correction results in large-scale processing, providing non-professional users with an efficient and convenient image processing tool. Attached Figure Description

[0033] Figure 1 This is a flowchart of a fully automatic atmospheric topographic radiometric correction method for high-resolution remote sensing images according to the present invention;

[0034] Figure 2 This is a schematic diagram illustrating the principle of atmospheric correction in this invention;

[0035] Figure 3This is a schematic diagram showing the comparison between remote sensing images of the same area before and after atmospheric correction and topographic radiation correction using the method of this invention. Detailed Implementation

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0037] Reference Figure 1 As shown, the implementation process of the fully automated atmospheric topographic radiometric correction method for high-resolution remote sensing imagery provided by this invention lies in the core of realizing a complete chain from data input to correction result output through a fully automated and intelligent discrimination processing framework. The entire process requires no manual intervention and supports batch operations, specifically including:

[0038] Step 1, Framework Initialization and Data Preparation: Input high-resolution image data, and automatically obtain the corresponding digital elevation model (DEM) data and atmospheric parameters;

[0039] The system first initiates a fully automated batch processing framework. Once the user or scheduler inputs the high-resolution satellite imagery data to be processed, the framework runs automatically and completes initialization. Its primary task is to read image metadata, including imaging time, sensor type, and geographic location information. Based on this information, the system automatically retrieves the corresponding digital elevation model (DEM) data from a preset data source (such as NASA's SRTM database) and prepares or calculates the necessary atmospheric parameters for subsequent atmospheric correction modules (e.g., obtaining atmospheric profile information from real-time meteorological data or climate models). This step provides a unified data foundation and calling environment for all subsequent correction modules.

[0040] Step 2, Atmospheric Correction: Perform automatic atmospheric correction on the high-resolution image based on the 6S radiative transfer model to obtain atmospherically corrected data;

[0041] A schematic diagram of the atmospheric correction principle is shown below. Figure 2 As shown, once the data is ready, the system automatically calls and executes the atmospheric correction module. The core of this module is based on the 6S radiative transfer model. It dynamically constructs the input scene of the 6S model using the image metadata, DEM, and atmospheric parameters acquired in the first stage, accurately simulating the complete transmission process of solar radiation from the top of the atmosphere to the Earth's surface and then to the sensor, taking into account the scattering and absorption of atmospheric molecules and aerosols, as well as path radiation effects. The module can automatically perform surface reflectance inversion, effectively removing atmospheric interference. In particular, this module integrates anomaly pixel detection, which can automatically identify and mark areas affected by clouds, fog, snow, etc., during processing, avoiding the impact of these noises on the overall correction accuracy, and finally outputting atmospherically corrected apparent reflectance data.

[0042] Step 3, Terrain Identification and Decision: Based on the digital elevation model (DEM), terrain feature parameters are automatically calculated, and the terrain complexity is analyzed through a pre-trained discrimination model to output the discrimination probability;

[0043] This is a key aspect demonstrating the intelligence of this invention. The system does not mechanically perform the same follow-up steps on all images; instead, it incorporates built-in terrain feature recognition and decision-making logic. Specifically, the system uses the acquired DEM data to perform terrain analysis on the image-covered area, automatically calculating parameters such as slope and aspect. Then, it uses algorithms to calculate indicators to determine whether the area belongs to a gently sloping terrain such as a "plain" or a complex terrain such as "mountains and hills."

[0044] Average slope Combined with the coefficient of variation (CV) of illumination as a display statistical feature and spatial feature (texture of slope map, illumination map), a statistical branch is constructed (coefficient of variation of illumination CV, average slope). Sun zenith angle ) and visual branch (input 256×256 slice image, first channel: slope map, second channel: The neural network, with the illumination map as two parallel input branches, concatenates the output features of the two branches and outputs the final terrain radiation correction necessity probability P through a fully connected layer.

[0045] Among them, average slope The calculation steps are as follows:

[0046] For each cell in the image DEM data ( , ), using 3 Window calculation of slope (in degrees). The third-order inverse distance squared weighting method is typically used.

[0047] ,

[0048] Where, gradient and The calculation method is as follows:

[0049] ,

[0050] ,

[0051] Note: arrive 3 The pixel elevation value within the window; i and j represent the row and column coordinate indices of the image DEM matrix, i representing the row number and j representing the column number; x represents the horizontal spatial distance (east-west direction); y represents the vertical spatial distance (north-south direction); z represents the elevation value of a single pixel; This represents the rate of change of elevation in the x-direction; This represents the rate of change of elevation in the y-direction. Indicates the size of a single cell;

[0052] Calculate the average slope of the entire map :

[0053] ,

[0054] In the formula, This represents the total number of rows in the image. This represents the total number of columns in the image.

[0055] The steps for calculating CV are as follows:

[0056] First, calculate the cosine of the incident angle. ), For the local solar incidence angle:

[0057] ,

[0058] In the formula, The zenith angle of the sun. For terrain slope, The azimuth of the sun. The slope of the terrain;

[0059] Next, the coefficient of variation is calculated. :

[0060] ,

[0061] In the formula, The cosine value of the incident angle standard deviation The cosine value of the incident angle The mean.

[0062] Based on DEM data, a combined matrix image of slope map and illumination map can be directly obtained. H and W represent the height and width dimensions of the combined matrix image, respectively; a statistical index vector is constructed based on the obtained parameters: , = ;

[0063] Using convolutional neural networks Extracting the combined matrix image Spatial feature vectors :

[0064] ,

[0065] in, This is global average pooling. Conv represents the activation function, and Conv represents the convolution operation. (Convolutional Neural Network) It can be any network from common existing technologies, which will not be elaborated here.

[0066] Using multilayer perceptron Statistical indicator vector Mapping to a high-dimensional feature space To make it compatible with visual features:

[0067] ,

[0068] in, This is the weight matrix. For bias vectors, Use the Sigmoid activation function;

[0069] The two sets of features are concatenated and passed through a classifier. Output the discrimination probability P:

[0070] ;

[0071] in, Represents vector operations. It is the Sigmoid activation function. For the weights of the fusion layer, For the bias of the fusion layer, For the weights of the output layer, This is the bias for the output layer.

[0072] Network output probability As used as the discrimination index, the terrain type is shown in Table 1:

[0073] Table 1

[0074]

[0075] Step 4: Adaptive selection process based on the discrimination probability value: If the probability value is less than the first threshold of 0.1, it is determined to be flat terrain (such as flat land). Then, the atmospherically corrected data is output; if the probability value is greater than or equal to the first threshold of 0.1, it is determined to be slightly undulating terrain or complex terrain (such as mountainous areas). Then, the VECA model is called to perform topographic radiation correction on the atmospherically corrected data, and the final surface reflectance product is output.

[0076] Specifically, if the terrain is determined to be slightly undulating, then topographic radiation correction is performed on the atmospherically corrected data based on the VECA model to obtain the final correction result.

[0077] If the area is identified as having complex terrain, the system automatically triggers a joint correction process. At this point, the atmospherically corrected data obtained in step 2, along with the high-precision DEM data, are input into the topographic radiation correction module. This module is based on the VECA model, and its principle fully considers the coupling relationship between vegetation cover and terrain. The module calculates the location-dependent solar incidence angle and terrain shading factor for each pixel, and quantifies the moderating effect of vegetation on topographic radiation distortion, thereby accurately compensating for and normalizing the unevenness of radiation brightness caused by differences in slope aspect and gradient in the image. This step aims to eliminate radiation distortion introduced by terrain undulations, ensuring that similar land features have consistent radiation performance in different terrain locations.

[0078] Step 5: Results integration and output.

[0079] For imagery with only atmospheric correction, the system directly processes and outputs the final apparent reflectance product. For imagery with combined correction, the system integrates the results processed by the topographic radiation correction module into the final surface reflectance product. The entire process is fully automated; the system can automatically perform format packaging, quality checks, and generate processing reports. Figure 3 As shown, after the image is processed by this method, the spectral characteristics of ground features are closer to reality after eliminating atmospheric effects and topographic distortion. The overall contrast and radiometric consistency of the image are significantly improved, and the details of the shadow areas are revealed, laying a reliable foundation for subsequent high-precision ground feature classification and quantitative remote sensing inversion.

[0080] By executing the above steps in sequence, this invention achieves a high degree of automation, intelligence, and batch processing of high-resolution image radiometric correction, significantly improving processing efficiency and accuracy, and reducing reliance on professional manual operation.

[0081] ENVI is a professional remote sensing image processing software widely used internationally in the field of remote sensing science. Table 2 compares the statistical indicators of the same batch of GF-2 multispectral images processed by this invention and ENVI software.

[0082] Table 2

[0083]

[0084] Table 2 shows the mean absolute error (MAE), root mean square error (RMSE), correlation coefficient (R), and coefficient of determination. As can be seen from the table, compared with ENVI, the method of this invention exhibits superior correction performance in all four bands: blue, green, red, and near-infrared. The mean absolute error (MAE) and root mean square error (RMSE) are generally reduced, while the correlation coefficient (R) and coefficient of determination (R²) are significantly improved, especially in the near-infrared band, where R² increases from 0.921 to 0.935. These data fully verify the high accuracy and reliability of the method of this invention in the joint correction of atmospheric and topographic radiation, demonstrating its significant advantages in improving image radiometric consistency and spectral accuracy.

[0085] The topographic radiation correction model used in this device is the VECA model. The VECA model and other models are compared before and after topographic radiation correction. The correlation coefficients between the radiance of the GF-2 remote sensing images and those of the GF-2 remote sensing images are shown in Table 3:

[0086] Table 3

[0087]

[0088] As can be seen from the table, before correction, radiance and cos t showed a significant positive correlation (e.g., the correlation coefficient for band 4 in image 2 reached 0.75), indicating that the radiation distortion caused by topographic relief was quite severe. After applying traditional SCS+C correction, C correction, and the VECA correction integrated in this invention, the correlation was significantly reduced or even turned into a weak negative correlation. Among them, the absolute values ​​of the correlation for each band were generally smaller after VECA correction (e.g., 0.09 for band 4 in image 2), indicating that it can more effectively eliminate the radiation differences caused by topography, verifying the superiority of the VECA model in topographic radiation compensation and the effectiveness of this method in topographic adaptive correction.

[0089] On the other hand, the present invention provides a fully automatic atmospheric topographic radiometric correction device for high-resolution remote sensing images, which includes various modules capable of implementing the various steps of the aforementioned method, specifically including:

[0090] The input module is used to input high-resolution remote sensing image data and automatically obtain the corresponding digital elevation model (DEM) data and atmospheric parameters.

[0091] The atmospheric correction module is used to perform automatic atmospheric correction on the high-resolution remote sensing image data based on the 6S radiative transfer model to obtain atmospherically corrected data.

[0092] The discrimination module is used to automatically calculate terrain feature parameters based on the digital elevation model (DEM) data, analyze the terrain complexity through a pre-trained discrimination model, and output the discrimination probability. The pre-trained discrimination model is a two-branch neural network that includes a visual branch and a statistical branch.

[0093] The terrain correction module is used to output the atmospherically corrected data when the discrimination probability value is less than a first threshold; if the discrimination probability value is greater than or equal to the first threshold, the VECA model is called to perform terrain radiation correction on the atmospherically corrected data and output the final surface reflectance product.

[0094] Thirdly, the present invention provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned fully automatic atmospheric topographic radiometric correction method for high-resolution remote sensing images.

[0095] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned method for fully automatic atmospheric and topographic radiometric correction of high-resolution remote sensing images.

[0096] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A high-resolution remote sensing image full-automatic atmospheric terrain radiation correction method, characterized in that, The method includes: Step 1: Input high-resolution remote sensing image data. Based on the metadata information of the high-resolution remote sensing image data, automatically retrieve and obtain the digital elevation model (DEM) data of the image coverage area from the preset external data source; and automatically extract or invert the atmospheric parameters required for 6S radiative transfer model calculation from the meteorological database or climate model based on the imaging time and geographical location. Step 2: Perform automatic atmospheric correction on the high-resolution remote sensing image data based on the 6S radiative transfer model to obtain atmospherically corrected data; Step 3: Automatically calculate terrain feature parameters based on the Digital Elevation Model (DEM) data. These parameters include: the average slope of the image coverage area calculated from the DEM data, and the coefficient of variation of the incident angle cosine calculated based on the solar-terrain geometric relationship. Then, analyze the terrain complexity using a pre-trained discriminant model and output a discriminant probability value. The pre-trained discriminant model is a two-branch neural network including a visual branch and a statistical branch. The input to the visual branch is a combined matrix image composed of a slope map and an illumination map generated from the DEM data, used to extract spatial texture features. The input to the statistical branch is a feature vector composed of the average slope, the coefficient of variation, and the solar zenith angle. After fusing the features from the visual and statistical branches, a classifier outputs a discriminant probability value representing the necessity of terrain radiometric correction. Step 4: Set the first threshold and the second threshold, and make a hierarchical decision based on the discrimination probability values: If the discrimination probability value is less than the first threshold, it is determined to be flat terrain, and terrain radiation correction is not performed; instead, atmospheric correction data is output. If the discrimination probability value is greater than or equal to the first threshold and less than the second threshold, it is determined to be a slightly undulating terrain. The VECA model is called to perform terrain radiation correction on the atmospheric correction data to obtain and output the final correction result. If the discrimination probability value is greater than or equal to the second threshold, it is determined to be complex terrain. The system automatically triggers the joint correction process, inputs the atmospheric correction data and the digital elevation model (DEM) data into the VECA model to perform terrain radiation correction, and obtains and outputs the final surface reflectance product.

2. The fully automatic atmospheric topographic radiometric correction method for high-resolution remote sensing images according to claim 1, characterized in that, The automatic atmospheric correction process in step 2 also includes: before or during the 6S radiative transfer model calculation, automatically detecting cloud, snow and thick haze coverage areas in the image and marking or masking them to avoid abnormal pixels from negatively affecting atmospheric parameter inversion and overall correction accuracy.

3. The fully automatic atmospheric topographic radiometric correction method for high-resolution remote sensing images according to claim 1, characterized in that, The first threshold is 0.1, and the second threshold is 0.

5.

4. A fully automatic atmospheric topographic radiometric correction device for high-resolution remote sensing images, used to implement the method described in any one of claims 1-3, characterized in that, include: The input module is used to input high-resolution remote sensing image data and automatically obtain the corresponding digital elevation model (DEM) data and atmospheric parameters. The atmospheric correction module is used to perform automatic atmospheric correction on the high-resolution remote sensing image data based on the 6S radiative transfer model to obtain atmospherically corrected data. The discrimination module is used to automatically calculate terrain feature parameters based on the digital elevation model (DEM) data, analyze the terrain complexity through a pre-trained discrimination model, and output a discrimination probability value. The pre-trained discrimination model is a two-branch neural network that includes a visual branch and a statistical branch. The terrain correction module is used to output the atmospheric correction data when the discrimination probability value is less than a first threshold. If the discrimination probability value is greater than or equal to the first threshold and less than the second threshold, the VECA model is invoked to perform topographic radiation correction on the atmospherically corrected data to obtain and output the final correction result. If the discrimination probability value is greater than or equal to the second threshold, the system automatically triggers the joint correction process, inputting the atmospheric corrected data and the digital elevation model (DEM) data into the VECA model to perform topographic radiation correction, and obtaining and outputting the final surface reflectance product.

5. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When one or more programs are executed by the one or more processors, the one or more processors implement the fully automatic atmospheric topographic radiometric correction method for high-resolution remote sensing images as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, enable the processor to implement the fully automatic atmospheric and topographic radiometric correction method for high-resolution remote sensing images as described in any one of claims 1-3.

Images