Multi-spatial resolution remote sensing simulation and error influence stripping method

Abstract

The application discloses a multi-spatial resolution remote sensing simulation and error influence stripping method, and relates to the technical fields of remote sensing image processing, quantitative remote sensing and environmental modeling. Firstly, a multi-scale remote sensing image is simulated by combining a modulation transfer function (MTF) and mean value interpolation; secondly, a multiple regression model is constructed to explain model errors by image quality indexes, so that the interference of image quality on analysis results is stripped, and normalized residual errors caused by changes in spatial resolution are obtained; finally, a sliding window-based RSI and "Median+IQR" scoring method is proposed to identify stable resolution intervals and perform next-step analysis, thereby guiding the selection of actual modeling scales. The method has high universality and scale adaptability, and is suitable for pre-scale analysis and optimized modeling in various remote sensing quantitative analysis tasks such as land cover, vegetation parameters, urban heat island, water body monitoring and the like.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing image processing and quantitative remote sensing analysis, and particularly relates to a method for multi-spatial resolution remote sensing simulation and error influence removal. Background Technology

[0002] Remote sensing data, as an important means of acquiring spatial information, plays a crucial role in fields such as environmental monitoring, land use classification, vegetation cover estimation, and water quality retrieval. However, the choice of spatial resolution for remote sensing images often faces a fundamental trade-off: while high-resolution images can provide richer details about ground features, they come with a huge amount of data and computational burden; while low-resolution images improve processing efficiency, they inevitably lose image details and introduce mixed pixel errors. This resolution-dependent effect can significantly affect the modeling feature extraction, prediction accuracy, and even the physical interpretability of results in downstream analysis tasks. Therefore, in various research settings such as environmental monitoring and resource surveys, it is necessary to systematically introduce the impact analysis of spatial scale on research results to determine the optimal observation scale.

[0003] Current research on the impact of spatial scale mainly employs empirical comparative analysis methods, assessing the scale effect by directly comparing differences in results at different resolutions. This approach has significant limitations: first, it cannot distinguish between the independent contribution of resolution change itself (i.e., the scale effect) and the resulting image quality degradation to the results; second, the lack of standardized resolution simulation procedures makes direct comparison between different studies difficult; and most importantly, the coupling effect between image quality degradation and modeling errors can interfere with the assessment of the true scale effect. Summary of the Invention

[0004] To address the above technical problems, this invention provides a method for multi-spatial-resolution remote sensing simulation and error removal. First, a high-fidelity simulated multi-resolution image sequence is constructed. Then, the "image damage effect" in the task error is removed by using image quality indicators. Finally, an independent residual sequence reflecting the ontological effect of spatial resolution is formed, thereby providing a unified and reliable scale research basis for various remote sensing analysis tasks.

[0005] The technical solution of this invention is:

[0006] A method for multi-spatial resolution remote sensing simulation and error removal includes the following steps:

[0007] S1: Blur the original high-resolution remote sensing images to construct a multi-spatial-scale remote sensing image sequence;

[0008] S2: Apply the inversion model to the remote sensing image sequence and quantify its prediction error; at the same time, introduce image quality indicators to characterize the error sources affecting reconstruction, and separate the degradation error in the image reconstruction process from the total prediction error, thereby obtaining the impact of scale on the inversion model.

[0009] S3: Set a sliding window, calculate the residual sensitivity index (RSI) within the window, identify the sensitivity variation of the inversion variables of the inversion model to resolution, and then determine the optimal spatial scale range by combining the principles of minimum error and stability.

[0010] Furthermore, step S1 specifically includes the following steps:

[0011] S11: Using remote sensing images with a spatial resolution of R0 as raw data, perform radiometric correction, atmospheric correction, and geometric correction;

[0012] S12: Based on the modulation transfer function (MTF) of the remote sensing sensor to simulate the spatial degradation process of the image, the image obtained in S11 is blurred by performing a blur convolution with an equivalent Gaussian kernel to obtain a blurred image; the formula for the standard deviation of the Gaussian kernel is as follows:

[0013] ;

[0014] in, The standard deviation of the equivalent Gaussian convolution kernel. This represents the modulation transfer function value of the remote sensing sensor used at the Nyquist frequency. This represents the original image spatial resolution.

[0015] The formula for Gaussian standard deviation is as follows:

[0016] ;

[0017] The formula for fuzzy convolution is as follows:

[0018] ;

[0019] ;

[0020] in, For the blurred image pixel value, For the original image Pixel value, Here, k represents the Gaussian convolution kernel weights, and k is the kernel radius. , , These represent the kernel index offsets;

[0021] S13: Resample the blurred image to multiple target resolutions to obtain a multi-spatial-scale remote sensing image sequence.

[0022] Furthermore, step S2 specifically includes the following steps:

[0023] S21: For each scale of remote sensing image, call the remote sensing analysis model and output parameter estimates;

[0024] S22: Will be composed of high-resolution remote sensing images The obtained output is used as the pseudo-true value. The error index corresponding to all prediction results at each spatial resolution is calculated and expressed as the mean absolute percentage error (MAPE).

[0025] ;

[0026] in, The number of inversion values ​​at a certain scale. For the i-th pseudo-true value, This is the i-th inversion value;

[0027] S23: Calculate the peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and spectral angle (SA) between the remote sensing image sequence and the original image sequence at each scale to characterize changes in image quality.

[0028] The formula for calculating PSNR is as follows:

[0029] ;

[0030] ;

[0031] in, Mean square error, This represents the maximum possible pixel value in the image. and The original image and images at various scales are located at... Pixel value at that location, and This represents the number of rows and columns of the image;

[0032] The formula for calculating SSIM is as follows:

[0033] ;

[0034] in, and Represents the local mean of the original image and images at various scales. and Represents the local standard deviation of the original image and images at various scales. For covariance, It is a constant;

[0035] The formula for calculating SA is as follows:

[0036] ;

[0037] in, , It is the dot product of two vectors; Let be the magnitude of the vector; and These are spectral vectors, each element... or A value representing the spectral reflectance or radiance of a wavelength band;

[0038] S24: Construct a multivariate regression model and extract the residuals as error components other than image quality:

[0039] ;

[0040] in, This represents the expected error value predicted based on image quality metrics at the i-th resolution scale. , , , representing the peak signal-to-noise ratio, structural similarity, and spectral angle at the i-th resolution scale, respectively, and a, b, c, and d are the coefficients of the regression model;

[0041] ;

[0042] in, It is the true, total prediction error observed at the i-th resolution scale. It is the residual at the i-th resolution scale;

[0043] S25: Normalize the residuals to the [0,1] interval and plot the residual sensitivity curve as a function of resolution;

[0044] ;

[0045] in, The normalized residuals It is the minimum residual value across all scales. It is the maximum residual value across all scales.

[0046] Furthermore, step S3 specifically includes the following steps:

[0047] S31: Using each scale point as the center, set a fixed-size sliding window, and calculate the residual sensitivity index (RSI) and the comprehensive score based on Median + IQR within the window:

[0048] ;

[0049] ;

[0050] in, Representing the The overall score of the sliding window Represents the set of normalized residuals within the sliding window. This represents the median of the normalized residuals within the sliding window. The interquartile range represents the normalized residuals within the sliding window;

[0051] S32: According to and Constructing a dual-criteria decision-making strategy, firstly based on The numerical values ​​determine a set of candidate windows such that the average error level of the candidate intervals is low, and then a selection is made from the candidate windows. The smallest window.

[0052] Furthermore, in step S21, the remote sensing analysis model is a regression, classification, or other estimation algorithm.

[0053] Furthermore, in step S21, the remote sensing analysis model can be a prediction model for the Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI), Land Surface Temperature (Ts), or water quality parameters such as chlorophyll aChl-a.

[0054] Furthermore, the Residual Sensitivity Index (RSI) is defined as the mean of the normalized residual curves within a window of a specified size, used to quantify the independent effect of resolution on model error.

[0055] Furthermore, the optimal resolution range refers to the lowest RSI and the lowest score based on Median+IQR, that is, finding a balance point that simultaneously satisfies small deviation and small fluctuation.

[0056] Furthermore, in S13, resampling selects any interpolation method to reconstruct the original image into multiple spatial resolution scales.

[0057] Furthermore, in S13, the target resolution is any equally spaced resolution interval required for the study.

[0058] The beneficial effects of this invention are as follows:

[0059] (1) An integrated process of “simulation-stripping-analysis” is proposed, and an image quality stripping method is introduced to analyze the sources of remote sensing modeling errors, which is supported by a theoretical mechanism;

[0060] (2) The residual sensitivity index RSI and the scoring criteria based on “Median + IQR” are proposed, which have the ability to evaluate resolution in a quantifiable and comparable manner.

[0061] (3) The sliding window evaluation strategy enables automatic identification of error stationary intervals and is applicable to multi-parameter, full-scale data;

[0062] (4) It supports flexible nesting of different remote sensing tasks, data types and model forms, and is suitable for application in most remote sensing quantitative analysis processes. Attached Figure Description

[0063] Figure 1 This is a technical flowchart of the present invention;

[0064] Figure 2 This is a trend image of the PSNR of pure water as a function of spatial resolution in an embodiment of the present invention;

[0065] Figure 3 This is a trend image of the SSIM of pure water as a function of spatial resolution in an embodiment of the present invention;

[0066] Figure 4 This is a trend image of SA (sulfuric acid) in pure water as a function of spatial resolution, according to an embodiment of the present invention.

[0067] Figure 5 This is a trend image of MAPE variation with spatial resolution for Chl-a in this embodiment of the invention;

[0068] Figure 6 This is a trend image of the normalized residual of Chl-a as a function of spatial resolution in an embodiment of the present invention;

[0069] Figure 7 This is a sliding window score chart based on "Median + IQR" in an embodiment of the present invention. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, this invention adopts the following technical solution.

[0071] This invention provides a method for multi-spatial resolution remote sensing simulation and error removal, comprising the following steps:

[0072] S1: Blur the original high-resolution remote sensing image (it can be based on modulation transfer function (MTF) blurring, but is not limited to this) to make the output image closer to the data acquired by the real remote sensing imaging system in terms of visual and spectral characteristics, and then construct a remote sensing image sequence with different spatial resolutions.

[0073] S2: Apply the existing inversion model (which can be a water quality parameter inversion model, but is not limited to it) to images at multiple spatial scales and quantify its prediction error; at the same time, introduce image quality indicators to characterize the error sources of image reconstruction, and separate the degradation error in the image reconstruction process from the total prediction error, thereby obtaining the impact of scale on the inversion model.

[0074] S3: Set a sliding window, calculate the residual sensitivity index (RSI) within the window, identify the sensitivity variation of the inversion variable to resolution, and then determine the optimal spatial scale range by combining the principles of minimum error and stability.

[0075] Furthermore, step S1 specifically includes the following steps:

[0076] S11: Using remote sensing images with a spatial resolution of R0 as raw data, perform radiometric correction, atmospheric correction, and geometric correction;

[0077] S12: Based on the modulation transfer function (MTF) of the remote sensing sensor to simulate the spatial degradation process of the image, the image obtained in S11 is blurred using an equivalent Gaussian convolution kernel. The standard deviation of the Gaussian kernel can be determined based on the sensor's MTF parameters and the original image resolution to simulate the spatial degradation characteristics of the remote sensing imaging system. The MTF parameters can be obtained through sensor design documents, experimental measurements, or on-orbit calibration, and are preferably the MTF values ​​of the sensor at the Nyquist frequency. .

[0078] The formula for the Gaussian kernel standard deviation is as follows:

[0079] ;

[0080] in, The standard deviation of the equivalent Gaussian convolution kernel (unit: m). This represents the modulation transfer function value (unitless, 0~1) of the remote sensing sensor used at the Nyquist frequency. The original image spatial resolution (unit: m).

[0081] The formula for Gaussian standard deviation (in pixels) is as follows:

[0082] ;

[0083] The formula for fuzzy convolution is as follows:

[0084] ;

[0085] ;

[0086] in, For the blurred image pixel value, For the original image Pixel value, Here, k represents the Gaussian convolution kernel weights, and k is the kernel radius, typically set to 3. , where m and n represent the kernel index offsets respectively.

[0087] S13: Resample the blurred image to multiple target resolutions to obtain a multi-scale remote sensing image sequence.

[0088] Furthermore, step S2 specifically includes the following steps:

[0089] S21: For each scale of remote sensing image, call the remote sensing analysis model and output parameter estimates;

[0090] S22: Will be composed of high-resolution images The obtained output is used as the pseudo-true value. The error index corresponding to all prediction results at each spatial resolution is calculated and expressed as the mean absolute percentage error (MAPE).

[0091] ;

[0092] Where N is the number of inversion values ​​at a certain scale. For the i-th pseudo-true value, Let be the i-th inversion value.

[0093] S23: Calculate the peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and spectral angle (SA) values ​​between the image at each scale and the original image to characterize image quality changes.

[0094] The formula for calculating PSNR is as follows:

[0095] ;

[0096] ;

[0097] in, Mean square error, This represents the maximum possible pixel value in the image. and The original image and images at various scales are located at... The pixel value at that location, where m and n are the number of rows and columns of the image;

[0098] The formula for calculating SSIM is as follows:

[0099] ;

[0100] in, and Represents the local mean of the original image and images at various scales. and Represents the local standard deviation of the original image and images at various scales. For covariance, It is a constant;

[0101] The formula for calculating SA is as follows:

[0102] ;

[0103] in, and These are spectral vectors, Each element or Represents the spectral reflectance or radiance value for a specific wavelength band. It is the dot product of two vectors; Let be the magnitude of the vector; The range is from 0° to 90°, and the smaller the angle, the higher the spectral similarity.

[0104] S24: Construct a multivariate regression model and extract the residuals as error components other than image quality:

[0105] ;

[0106] in, This represents the expected error value predicted based on image quality metrics at the i-th resolution scale. , , denoted as peak signal-to-noise ratio, structural similarity, and spectral angle at the i-th resolution scale, respectively, and a, b, c, and d are the coefficients of the regression model.

[0107] ;

[0108] in, It is the true, total prediction error observed at the i-th resolution scale. It is the residual at the i-th resolution scale.

[0109] S25: Normalize the residuals to the [0,1] interval and plot the residual sensitivity curve as a function of resolution;

[0110] ;

[0111] in, The normalized residuals It is the minimum residual value across all scales. It is the maximum residual value across all scales.

[0112] Furthermore, step S3 specifically includes the following steps:

[0113] S31: Using each scale point as the center, set a fixed-size sliding window, calculate the residual sensitivity index (RSI) and the comprehensive score based on "Median + IQR" within the window:

[0114] ;

[0115] ;

[0116] in, Representing the The overall score of the sliding window Represents the set of normalized residuals within the sliding window. This represents the median of the normalized residuals within the sliding window. This represents the interquartile range of the normalized residuals within the sliding window.

[0117] S32: According to and based on "Median + IQR" Constructing a dual-criteria decision-making strategy, firstly based on The numerical values ​​determine a set of candidate windows that ensure a low average error level for the candidate intervals, and then a selection is made from these candidate windows. The smallest window. This criterion combines the advantages of low average error and high stability.

[0118] Furthermore, in step S21, the remote sensing analysis model is a regression, classification, or other estimation algorithm.

[0119] Furthermore, in step S21, the remote sensing analysis model can be a prediction model of the normalized vegetation index (NDVI), leaf area index (LAI), land surface temperature (Ts), or water quality parameters such as chlorophyll a (Chl-a).

[0120] Furthermore, the Residual Sensitivity Index (RSI) is defined as the mean of the normalized residual curves within a window of a specified size, used to quantify the independent effect of resolution on model error.

[0121] Furthermore, the optimal resolution range refers to the lowest RSI and the lowest Score based on "Median + IQR", that is, finding a balance point that simultaneously satisfies "small deviation" and "small fluctuation".

[0122] Furthermore, in S13, resampling can select any interpolation method to reconstruct the original image at multiple spatial resolution scales.

[0123] Furthermore, in S13, the target resolution is any equally spaced resolution interval required for the study.

[0124] like Figure 1 As shown, this invention proposes a multi-spatial-resolution remote sensing simulation and error impact removal method. Taking the Chl-a inversion task for small and medium-sized lakes as an example, its core idea is: based on existing high-resolution remote sensing images, simulate and generate remote sensing image sequences covering multiple spatial resolutions; by constructing a mechanism to separate image quality from model error, quantitatively identify the independent impact of spatial resolution on model performance, thereby selecting the most suitable modeling resolution range.

[0125] S1: Based on the system's MTF characteristics, a spatial degradation model is constructed. Gaussian blurring and resampling methods are used to generate remote sensing image sequences with different spatial resolutions to ensure that images at each scale have the consistency of the real imaging process.

[0126] S2: Construct a consistent Chl-a inversion model on multi-scale images and quantify its prediction error. At the same time, introduce image quality indicators (such as PSNR, SSIM and SA) to model the source of error. Use regression methods to separate "image degradation error" from the total error to obtain a purer "scale effect" component.

[0127] S3: Normalize and trend-analyze the residuals at each scale to identify the sensitivity variation of different water quality parameters to resolution. Then, based on the principles of minimum error and stability, determine the optimal spatial scale range to provide a basis for the selection of resolution in actual remote sensing monitoring.

[0128] A further preferred embodiment of the present invention is that step S1 specifically includes the following steps:

[0129] S11: Using airborne hyperspectral remote sensing images with a spatial resolution of 0.75m as raw data, perform data preprocessing such as radiometric correction, atmospheric correction, geometric correction, and band resampling;

[0130] S12: Based on the Sentinel-2 MSI modulation transfer function (MTF) model, the spatial blurring characteristics of the imaging system are simulated. The image is processed using a Gaussian convolution kernel, where the standard deviation of the kernel function is... Based on the system response characteristics of the Sentinel-2 MSI sensor in different bands, and using the MTF value of the sensor at the Nyquist frequency... Then, a two-dimensional Gaussian blur kernel is constructed, and a global convolution operation is performed on the original image to achieve spatial degradation processing, thereby obtaining a blurred image.

[0131] ;

[0132] S13: The blurred image is resampled to multiple target resolutions (1.5~120m, with an interval of 0.75m) using the mean interpolation method to obtain a multi-scale image sequence.

[0133] A further preferred embodiment of the present invention is that step S2 specifically includes the following steps:

[0134] S21: For each scale image, call the established Chl-a remote sensing inversion model;

[0135] S22: Use the output of the original high-resolution image model as the "pseudo-true value" and calculate the mean absolute percentage error (MAPE) for each scale.

[0136] S23: Calculate the PSNR, SSIM, and SA values ​​between the image at each scale and the original image to characterize changes in image quality.

[0137] S24: Construct a multivariate regression model (PSNR+SSIM+SA→MAPE) and extract the residuals as error components other than image quality;

[0138] S25: Normalize the residuals to the [0,1] interval and plot the residual sensitivity curve as a function of resolution.

[0139] A further preferred embodiment of the present invention is that step S3 specifically includes the following steps:

[0140] S31: Set up a sliding window and calculate the residual sensitivity index (RSI) and the comprehensive score based on "Median + IQR" within the window;

[0141] S32: Select the interval with the lowest score from the set of sets with low residual sensitivity indices as the recommended interval for optimal modeling resolution under this analysis task.

[0142] Example:

[0143] This embodiment uses a general scale simulation and error stripping analysis method for quantitative remote sensing proposed in this invention to select the optimal spatial resolution range for remote sensing inversion of Chl-a concentration in Dianshan Lake and Yuandang Lake areas. First, 0.75m resolution airborne hyperspectral images acquired in June 2022 were selected as the baseline data. Then, a Gaussian blur kernel was constructed based on the MTF characteristics of Sentinel-2 MSI to generate a simulated image sequence with 160 gradients from 0.75m to 120m (intervals of 0.75m). Experimental results show that: (1) Image quality indicators: PSNR and SSIM values ​​decrease exponentially with decreasing resolution, such as Figure 2 , Figure 3 As shown, the SA value increases exponentially, such as... Figure 4 As shown, it conforms to the expectations of the spatial degradation theory; (2) Inversion error characteristics: the MAPE of Chl-a concentration shows a linear increasing trend, as shown in the figure. Figure 5 As shown, the reduction in resolution directly leads to a decrease in inversion accuracy; (3) Error sensitivity analysis: as shown Figure 6 As shown, the normalized residual curve exhibits significant fluctuations throughout the entire interval, revealing the important influence of the scaling effect in this range. Finally, a sliding window with a width of 20 meters was used for analysis. First, candidate intervals with lower average errors were selected based on the Residual Sensitivity Index (RSI). Then, based on the principle of minimizing the comprehensive score (Score = Median + IQR), the interval of 18.2–38.2 meters was determined as the optimal spatial resolution interval. Figure 7 Black striped window.

[0144] It should be noted that the technical solution of the present invention is not limited to the specific embodiments described above. All technical modifications made according to the technical solution of the present invention fall within the protection scope of the present invention.

Claims

1. A method for multi-spatial resolution remote sensing simulation and error influence removal, characterized in that, Includes the following steps: S1: Blur the original high-resolution remote sensing images to construct a multi-spatial-scale remote sensing image sequence; S2: Apply the inversion model to the remote sensing image sequence and quantify its prediction error; at the same time, introduce image quality indicators to characterize the error sources affecting reconstruction, and separate the degradation error in the image reconstruction process from the total prediction error, thereby obtaining the impact of scale on the inversion model. S3: Set a sliding window, calculate the residual sensitivity index (RSI) within the window, identify the sensitivity variation of the inversion variables of the inversion model to resolution, and then determine the optimal spatial scale range by combining the principles of minimum error and stability. The Residual Sensitivity Index (RSI) is defined as the mean of the normalized residual curves within a window of a specified size, and is used to quantify the independent effect of resolution on model error.

2. The method for multi-spatial resolution remote sensing simulation and error influence removal according to claim 1, characterized in that, Step S1 specifically includes the following steps: S11: Using remote sensing images with a spatial resolution of R0 as raw data, perform radiometric correction, atmospheric correction, and geometric correction; S12: Based on the modulation transfer function (MTF) of the remote sensing sensor to simulate the spatial degradation process of the image, the image obtained in S11 is blurred by performing a blur convolution with an equivalent Gaussian kernel to obtain a blurred image; the formula for the standard deviation of the Gaussian kernel is as follows: ； in, The standard deviation of the equivalent Gaussian convolution kernel. This represents the modulation transfer function value of the remote sensing sensor used at the Nyquist frequency. This represents the original image spatial resolution. The formula for Gaussian standard deviation is as follows: ； The formula for fuzzy convolution is as follows: ； ； in, For the blurred image pixel value, For the original image Pixel value, Here, k represents the Gaussian convolution kernel weights, and k is the kernel radius, which is set to 3. , , These represent the kernel index offsets; S13: Resample the blurred image to multiple target resolutions to obtain a multi-spatial-scale remote sensing image sequence.

3. The method for multi-spatial resolution remote sensing simulation and error influence removal according to claim 2, characterized in that, Step S2 specifically includes the following steps: S21: For each scale of remote sensing image, call the remote sensing analysis model and output parameter estimates; S22: Will be composed of high-resolution remote sensing images The obtained output is used as the pseudo-true value. The error index corresponding to all prediction results at each spatial resolution is calculated and expressed as the mean absolute percentage error (MAPE). ； in, The number of inversion values ​​at a certain scale. For the i-th pseudo-true value, This is the i-th inversion value; S23: Calculate the peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and spectral angle (SA) between the remote sensing image sequence and the original image sequence at each scale to characterize changes in image quality. The formula for calculating PSNR is as follows: ； ； in, Mean square error, This represents the maximum possible pixel value in the image. and The original image and images at various scales are located at... Pixel value at that location, and This represents the number of rows and columns of the image; The formula for calculating SSIM is as follows: ； in, and Represents the local mean of the original image and images at various scales. and Represents the local standard deviation of the original image and images at various scales. For covariance, It is a constant; The formula for calculating SA is as follows: ； in, , It is the dot product of two vectors; Let be the magnitude of the vector; and These are spectral vectors, each element... or A value representing the spectral reflectance or radiance of a wavelength band; S24: Construct a multivariate regression model and extract the residuals as error components other than image quality: ； in, This represents the expected error value predicted based on image quality metrics at the i-th resolution scale. , , , representing the peak signal-to-noise ratio, structural similarity, and spectral angle at the i-th resolution scale, respectively, and a, b, c, and d are the coefficients of the regression model; ； in, It is the true, total prediction error observed at the i-th resolution scale. It is the residual at the i-th resolution scale; S25: Normalize the residuals to the [0,1] interval and plot the residual sensitivity curve as a function of resolution; ； in, The normalized residuals It is the minimum residual value across all scales. It is the maximum residual value across all scales.

4. The method for multi-spatial resolution remote sensing simulation and error influence removal according to claim 3, characterized in that, Step S3 specifically includes the following steps: S31: Using each scale point as the center, set a fixed-size sliding window, and calculate the residual sensitivity index (RSI) and the comprehensive score based on Median + IQR within the window: ； ； in, Representing the The overall score of the sliding window Represents the set of normalized residuals within the sliding window. This represents the median of the normalized residuals within the sliding window. The interquartile range represents the normalized residuals within the sliding window; S32: According to and Constructing a dual-criteria decision-making strategy, firstly based on The numerical values ​​determine a set of candidate windows such that the average error level of the candidate intervals is low, and then a selection is made from the candidate windows. The smallest window.

5. The method according to claim 3, characterized in that, In step S21, the remote sensing analysis model is a regression, classification, or other estimation algorithm.

6. The method according to claim 5, characterized in that, In step S21, the remote sensing analysis model is a prediction model for the Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI), Land Surface Temperature (Ts), or water quality parameters such as chlorophyll aChl-a.

7. The method according to claim 3, characterized in that, The optimal resolution range refers to the lowest RSI and the lowest score based on Median+IQR, that is, finding a balance point that simultaneously satisfies small deviation and small fluctuation.

8. The method according to claim 2, characterized in that, In S13, resampling selects any interpolation method to reconstruct the original image into multiple spatial resolution scales.

9. The method according to claim 2, characterized in that, In S13, the target resolution is any equally spaced resolution interval required for the study.

Images