Water and soil loss pattern spot landing system based on weak constraint of land utilization information

Abstract

A water and soil loss pattern spot landing system based on land utilization information weak constraint comprises a land utilization information acquisition module, a factor evaluation and analysis module, a time sequence analysis and pattern spot landing module and a data integration and visualization module. The factor evaluation and analysis module is used for calculating a water and soil loss erosion modulus and evaluating and analyzing influence factors, the time sequence analysis and pattern spot landing module is used for analyzing water and soil loss changes and generating landing pattern spots, and the data integration and visualization module is used for data integration and pattern spot visualization display. According to the water and soil loss pattern spot landing system based on the land utilization information weak constraint, an erosion modulus calculation algorithm based on data fusion and pixel unmixing is proposed to calculate an erosion modulus, and a water and soil loss pattern spot change analysis algorithm based on time sequence classification is proposed to analyze pattern spot change.

Description

Technical Field

[0001] The present invention relates to the technical fields of data fusion, pixel unmixing, and time series classification, and specifically to a soil and water loss patch landing system based on weak constraints of land use information. Background Art

[0002] Data fusion technology is a technology that integrates data from different sources to improve data quality and accuracy, aiming to solve the problems of insufficient information and poor accuracy of a single data source. This technology improves the spatial resolution and accuracy of soil and water loss monitoring by fusing multiple data sources such as remote sensing data, geographic information system data, and land use information. In the soil and water loss patch landing system based on weak constraints of land use information, data fusion technology is used to calculate the soil erosion modulus. By fusing different types of factors, including rainfall erosion factor, soil erodibility factor, slope factor, slope length factor, vegetation cover factor, soil and water conservation engineering measure factor, and tillage measure factor, the intensity and distribution of soil erosion can be obtained more comprehensively, thus ensuring that the generated soil and water loss patches can accurately reflect the actual erosion situation.

[0003] Pixel unmixing technology is a technology for extracting a single category from mixed pixels, aiming to solve the problem of information loss caused by pixel mixing in remote sensing data. In the soil and water loss patch landing system based on weak constraints of land use information, pixel unmixing technology is applied to calculate the erosion modulus. By unmixing the pixels in the remote sensing image, independent signals representing various land uses and soil and water loss conditions are extracted, so as to accurately obtain the physical parameters related to soil erosion, and further ensure the accuracy and reliability of the erosion modulus calculation.

[0004] Time series classification technology is a technology for pattern recognition and classification based on time series data, aiming to solve the problem of analyzing the change patterns of soil and water loss patches in different time periods. In the soil and water loss patch landing system based on weak constraints of land use information, time series classification technology is used to analyze the temporal changes of soil and water loss patches. By classifying and analyzing the dynamic monitoring data of soil and water loss patches over the years, the system can identify the long-term change trends of soil and water loss, distinguish stable areas from changing areas, and thus provide a scientific basis for soil and water loss control.

[0005] However, an existing soil and water loss patch landing system based on weak land use information constraints has the problem that the accuracy of soil and water loss patches is limited by a single data source, resulting in inaccurate calculation of soil erosion modulus and affecting the effect of soil and water loss monitoring. Secondly, due to the pixel mixing problem in remote sensing data, it is impossible to effectively extract single land use and soil and water loss information, resulting in blurred patch boundaries and misjudgment of erosion intensity. Finally, in the analysis of soil and water loss patch changes, there is a lack of dynamic analysis based on time series, and the long-term change trend cannot be accurately captured, resulting in insufficient evaluation of patch stability and affecting the guidance for soil and water loss control measures. Summary of the Invention

[0006] The purpose of the present invention is to provide a soil and water loss patch landing system based on weak land use information constraints, so as to solve the problems existing in the existing soil and water loss patch landing system based on weak land use information constraints as mentioned in the above background technology. The accuracy of soil and water loss patches is limited by a single data source, resulting in inaccurate calculation of soil erosion modulus and affecting the effect of soil and water loss monitoring. Secondly, due to the pixel mixing problem in remote sensing data, it is impossible to effectively extract single land use and soil and water loss information, resulting in blurred patch boundaries and misjudgment of erosion intensity. Finally, in the analysis of soil and water loss patch changes, there is a lack of dynamic analysis based on time series, and the long-term change trend cannot be accurately captured, resulting in insufficient evaluation of patch stability and affecting the guidance for soil and water loss control measures.

[0007] To achieve the above object, the present invention provides the following technical solutions: A soil and water loss patch landing system based on weakly constrained land use information, including a land use information collection module, a factor evaluation and analysis module, a time series analysis and patch landing module, and a data integration and visualization module, characterized in that: the land use information collection module is used to collect and integrate land use related data to ensure accurate land use basic information for subsequent analysis; the factor evaluation and analysis module includes an erosion modulus calculation unit and a factor classification and grading unit. The erosion modulus calculation unit proposes an erosion modulus calculation algorithm based on data fusion and pixel unmixing to accurately calculate the soil erosion modulus of each patch, ensuring that the calculation result of the erosion intensity can truly reflect the soil and water loss degree of different plots. The factor classification and grading unit is used to classify and grade each influencing factor to ensure that the soil and water loss factors are accurately mapped to the corresponding patches; the time series analysis and patch landing module includes a soil and water loss change analysis unit and a soil and water loss patch generation unit. The soil and water loss change analysis unit proposes a soil and water loss patch change analysis algorithm based on time series classification to analyze the patch changes to ensure accurate capture of the patch change trend. The soil and water loss patch generation unit is used to generate soil and water loss patches according to the analysis results to ensure the accuracy and clear boundary of the patch landing; the data integration and visualization module is used to integrate and visually display the processed data to ensure that users can intuitively obtain the soil and water loss analysis results and treatment suggestions.

[0008] Preferably, the land use information collection module collects and processes land use data, including land types, plot information, and land use data, to ensure accurate and efficient land use basic data for subsequent generation and analysis of soil and water loss patches.

[0009] Preferably, the factor evaluation and analysis module includes an erosion modulus calculation unit. The erosion modulus calculation unit proposes an erosion modulus calculation algorithm based on data fusion and pixel unmixing. By combining remote sensing data from different sources with land use information, pixel unmixing technology is used to improve data resolution, and data fusion methods are used to optimize the calculation accuracy of each factor to ensure accurate assessment of the soil erosion modulus of each patch and provide accurate basic data support for effective treatment of soil and water loss.

[0010] Preferably, the erosion modulus calculation algorithm based on data fusion and pixel unmixing is as follows: First, use data fusion to eliminate the uncertainty of a single data source and improve the reliability of the algorithm input parameters. Construct a pixel-level weighted data fusion algorithm to spatially align and fuse the band reflectance of the multispectral remote sensing image with the elevation information of the digital elevation model (DEM). Introduce land use vector data as a weak constraint condition, and optimize the fusion weight through spatial overlay analysis. The specific formula is expressed as:

[0011]

[0012] Among them, F i,j represents the comprehensive eigenvalue obtained after multi-source data fusion at the pixel position (i, j), w k represents the contribution degree of the k-th type of data source in the fusion process, D k,i,j represents the specific value of the k-th type of data source at the pixel position (i, j). The data sources include different band data of remote sensing images and DEM elevation information data. i represents the row index of the pixel in space, which is used to locate the vertical position of the pixel in the two-dimensional grid data. j represents the column index of the pixel in space, which is used to locate the horizontal position of the pixel in the two-dimensional grid data. n represents the total number of data sources, k represents the data source number index, which is used to distinguish different data sources, α represents the weak constraint adjustment factor used to adjust the contribution degree of land use information to the fusion result, L i,j represents the constraint effect of the land use type on the fusion result at the pixel position (i, j). By weakening the unreasonable areas through land use information, the fused data can retain spectral features and terrain features. Then, the pixel unmixing technology is used to separate the vegetation, soil, and water body endmember ratios in the mixed pixels, and the vegetation coverage is accurately extracted. Combining with the fused DEM data, the slope and slope length are calculated to provide dynamic parameters for the erosion modulus model. The specific formula for vegetation coverage unmixing is expressed as:

[0013]

[0014] Among them, V i,j represents the vegetation coverage of the pixel (i, j), R b,i,j represents the reflectance of band b at the pixel (i, j), S b represents the standard reflectance of the soil endmember in band b, V b represents the standard reflectance of the vegetation endmember in band b, m represents the total number of bands. The specific formula for improved nine-point slope length calculation is expressed as:

[0015]

[0016] Among them, L new represents the slope length of the pixel center point, h z represents the elevation value of the z-th pixel around the pixel center. z represents the index of the number of pixels around the pixel center, h center represents the elevation value of the center pixel, represents the horizontal distance between the z-th pixel and the center pixel, cos represents the cosine function, θ center represents the slope angle of the center pixel. The specific formula for slope calculation is expressed as:

[0017]

[0018] Among them, S new represents the improved slope factor of the pixel center point, represents the elevation gradient, which is calculated by the elevation difference between the central pixel and the surrounding pixels. Arctan represents the arctangent function. The selection of endmembers is constrained by land use types to avoid the interference of bare land and building areas on the unmixing results. Secondly, based on the framework of the Universal Soil Loss Equation (USLE) model, combined with the dynamic parameters of data fusion and pixel unmixing, the calculation of erosion modulus per pixel is realized, and the model coefficients are optimized using land use information. The erosion modulus formula is defined, a dynamic adjustment factor is introduced, and the erosion inhibition factor is adjusted according to land use types. The specific calculation formula is expressed as:

[0019] E = β·R·K·L new ·S new (1 - V)·P land

[0020] Among them, E represents the erosion modulus, β represents the dynamic calibration coefficient, R represents the rainfall erosivity factor, K represents the soil erodibility factor, V represents the vegetation coverage, and P land represents the land use inhibition factor, and P land is directly associated with the weak constraint of land use information. Through the dynamic assignment of vector plot attributes, the spatial differential correction of the erosion modulus is realized. Finally, the erosion modulus raster map is classified according to national standards and converted into vector polygons. Combining with the land use boundary, the spatial continuity of the polygons is optimized to meet the management requirements. The specific formula for the vector polygon optimization rule is expressed as:

[0021] If A patch < T and L land = cultivated land, then it is merged into the adjacent forest land

[0022] Among them, A patch represents the area of the vector polygon, T represents the area threshold, and L land represents the dominant land use type of the polygon. By constraining the vector generalization process with the land use boundary, it is ensured that the erosion polygons are consistent with the actual boundaries of farmland and forest land plots, improving the implementation of the results.

[0023] Preferably, the factor evaluation and analysis module includes a factor classification and grading unit. The factor classification and grading unit classifies and grades various factors, including rainfall erosion factor, soil erodibility factor, slope factor, slope length factor, vegetation coverage factor, soil and water conservation engineering measure factor, and tillage measure factor, to ensure that the factors related to soil erosion are accurately mapped onto the plot polygons for subsequent polygon analysis and processing.

[0024] Preferably, the time series analysis and map spot landing module includes a soil and water loss change analysis unit. The soil and water loss change analysis unit proposes an algorithm for analyzing the changes in soil and water loss map spots based on time series classification. By analyzing the time series data of soil and water loss map spots over the years, it identifies the change patterns and trends of the map spots, and combines with the land use change situation to ensure that the dynamic changes and stability of soil and water loss map spots can be accurately judged, providing a reliable basis for the long-term monitoring and treatment of soil and water loss.

[0025] Preferably, the algorithm for analyzing the changes in soil and water loss map spots based on time series classification is as follows: First, by separating the trend term and seasonal term in the time series, the long-term change law of soil and water loss is analyzed, providing a basis for change point detection. The specific formula is expressed as:

[0026]

[0027] where, T(t) represents the global trend of the time series, ε represents the basic intercept, η represents the linear slope, δ x represents the trend change amplitude of the x-th change point, t represents the time variable, X represents the total number of change points detected in the time series, τ x represents the time position of the x-th change point, representing the time point where the trend changes suddenly. I represents the indicator function. The I function is used to introduce a change in the trend after the change point. When the time t is greater than or equal to the change point τ x , the function value is 1, otherwise it is 0. The specific formula of the seasonal model is expressed as:

[0028]

[0029] where, S(t) represents the seasonal component of the time series at time t, representing the mathematical expression of the periodic change in the time series. n represents the order of the Fourier series, used to describe the complexity of the seasonal change. N represents the highest order of the Fourier series. a n represents the coefficient of the n-th order cosine term, used to describe the amplitude of the cosine component in the seasonal change. b n represents the coefficient of the n-th order sine term, used to describe the amplitude of the sine component in the seasonal change. represents the n-th order cosine term, used to capture the cosine fluctuation in the seasonal change. represents the n-th order sine term, used to capture the sine fluctuation in the seasonal change. P represents the seasonal period. The detection threshold of the change point τ is adjusted according to the land use type through the trend model to ensure that the change point is related to soil and water loss. Through the seasonal model: The Fourier coefficients a x and b n and b n, suppress seasonal noise unrelated to soil and water loss, and then identify the mutation points of soil and water loss in the time series, and segment them into stable subsequences to provide independent analysis units for classification. The specific formula is expressed as:

[0030] R(t) = V fused (t) - [T(t) + S(t)]

[0031] Among them, R(t) represents the residual value at time t, representing the difference between the observed value and the model predicted value, expressed as the pixel value after multi-source data fusion at time t. T(t) represents the trend term at time t, representing the long-term change trend of the time series. S(t) represents the seasonal term at time t, representing the periodic change in the time series. The specific formula for change point determination is expressed as:

[0032] BIC = -2ln(L) + ζln(T)

[0033] Among them, BIC represents the Bayesian Information Criterion, a statistic used to evaluate the goodness of fit of the model. ln(·) represents the logarithmic function. L represents the likelihood function, representing the fitting degree of the algorithm to the observed data. ζ represents the number of model parameters. T represents the total length of the time series.

[0034] Calculate the difference between the observed value and the model predicted value at each time point through residual analysis to identify abnormal points and change points. Calculate the change point determination through the results of residual analysis. Evaluate the rationality of different change point hypotheses through the Bayesian Information Criterion BIC, and select the optimal number and location of change points. ln(·) represents the natural logarithmic function. Secondly, through pixel unmixing and patch classification, solve the problem of mixed pixels, accurately distinguish soil and water loss patches from background ground objects, and improve the accuracy of classification boundaries. The specific formula for linear unmixing is expressed as:

[0035]

[0036] Among them, V mixed (t) represents the mixed pixel value at time t. f c represents the abundance of the c-th type of ground object, representing the proportion of the c-th type of ground object in the mixed pixel. V c (t) represents the pure spectral time series value of the c-th type of ground object at time t. ∈ represents the noise term, representing the residual part that the algorithm fails to explain. C represents the total number of ground object categories. The specific formula for dynamic time warping classification is expressed as:

[0037]

[0038] Among them, DWB(O, R) represents the dynamic time warping distance between time series O and R. O represents the time series to be classified, R represents the reference time series, w represents the alignment path, which represents the optimal alignment method between time series O and R. U represents the total length of the alignment path, and u represents the length index of the alignment path. represents the value of O corresponding to the u-th point in the alignment path w. represents the value of R corresponding to the u-th point in the alignment path w. represents minimizing the total distance of the alignment path w. By decomposing the spectral features of mixed pixels, the abundances of various ground object types are obtained to solve the mixed pixel problem. By non-linearly aligning time series, the similarity between the sequence to be classified and the reference sequence is measured to achieve high-precision classification. Finally, by fusing the prior information of land use, the patch boundaries are optimized to generate high-precision soil erosion patch maps. The specific formula for spatial probability fusion is expressed as follows:

[0039] P final (x, y) = λ · P class (x, y) + (1 - λ) · P landuse (x, y)

[0040] Among them, P final (x, y) represents the final probability value at the position (x, y), representing the probability that the position (x, y) belongs to the soil erosion patch. P class (x, y) represents the soil erosion probability obtained by dynamic time warping classification at the position (x, y). P landuse (x, y) represents the prior probability based on historical land use data at the position (x, y). λ represents the fusion weight. The specific formula for level set boundary optimization is expressed as follows:

[0041]

[0042] Among them, φ t (x, y) represents the level set function value at the position (x, y) at time t. φ t+1 (x, y) represents the level set function value at the position (x, y) at time t + 1. η represents the evolution rate parameter, which controls the step size of the level set function update. represents the curvature term of the level set function, representing the change in boundary curvature. represents the gradient of the level set function, representing the rate of change of the function value in space. represents the magnitude of the gradient of the level set function. κ represents the weight parameter, which is used to control the influence of the probability term on the boundary evolution. P finalRepresents the final probability value at the position (x, y). By data fusion, the spatio-temporal continuity is enhanced. Pixel unmixing solves the problem of mixed pixels. Using weak land use constraints, false detections are suppressed through probability fusion to ensure that the patch boundaries are consistent with the actual field. Combining temporal analysis and spatial optimization, the accurate localization of soil erosion patches is achieved.

[0043] Preferably, the temporal analysis and patch localization module includes a soil erosion patch generation unit. The soil erosion patch generation unit generates accurate soil erosion patches by combining different soil erosion factors with land use information according to the analysis results of the soil erosion change analysis unit, and performs smoothing and merging processing on the patches to ensure that the generated patch boundaries are clear and conform to the actual distribution of soil erosion.

[0044] Preferably, the data integration and visualization module integrates all the generated data and analysis results for visual display, ensuring that users can intuitively view the soil erosion status, erosion intensity, and patch changes, thereby providing effective decision-making support to help implement accurate soil erosion control measures.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] 1. The erosion modulus calculation unit proposes an erosion modulus calculation algorithm based on data fusion and pixel unmixing. First, the algorithm uses data fusion technology to effectively integrate the band reflectance information of multi-spectral remote sensing images and the elevation information of the digital elevation model (DEM), and combines land use vector data as weak constraint conditions to optimize the fusion weights, so as to eliminate the errors and uncertainties brought by a single data source. Through pixel-level weighted data fusion, the spatial matching and data consistency of various factors are ensured, enabling the erosion modulus calculation to accurately reflect the soil and water loss characteristics of different land types. In addition, introducing land use information as a spatial constraint not only enhances the physical meaning of data fusion but also effectively avoids the problem of class confusion caused by spectral mixing effects in remote sensing images, thereby improving the reliability of erosion modulus calculation. The fused data retains both the spectral information of the remote sensing image and the topographic features of the DEM, enabling the soil erosion modulus calculation to be optimized based on more complete geographical environment information, improving the spatial continuity and authenticity of the calculation results. Second, the algorithm uses pixel unmixing technology to deeply analyze remote sensing images, accurately separate the proportions of vegetation, soil, and water endmembers in mixed pixels, extract high-precision vegetation coverage information, and calculate slope and slope length in combination with DEM data, thereby optimizing the soil erosion modulus calculation model. The pixel unmixing process effectively reduces the errors caused by pixel mixing in remote sensing data, making the calculation of the vegetation coverage factor more accurate. Especially in areas with more mixed pixels, this technology can effectively improve the analysis accuracy of vegetation coverage information. When calculating slope length and slope in combination with DEM, an improved nine-point slope length calculation method is adopted to improve the calculation accuracy of the slope length factor and make the slope information more in line with the actual terrain conditions, providing more refined parameter support for soil erosion intensity assessment. In addition, during the pixel unmixing process, through the constraint of land use types, the interference of bare land and building areas on vegetation coverage calculation is effectively eliminated, avoiding misclassifying non-vegetation areas as high-vegetation coverage areas, thereby further improving the credibility of erosion modulus calculation. Finally, based on the dynamic parameters obtained from data fusion and pixel unmixing, the algorithm realizes per-pixel erosion modulus calculation within the framework of the Universal Soil Loss Equation (USLE) model, optimizes the model coefficients in combination with land use information, and introduces a dynamic adjustment factor, enabling the erosion modulus to adaptively adjust with changes in the spatial environment and land types. To meet management and application requirements, the algorithm also optimizes the vectorized erosion patches through land use boundaries, improves the spatial consistency of soil and water loss patches, makes their boundaries more in line with the actual plot morphology, and enhances the practicality and applicability of the results.

[0047] 2. The soil and water loss change analysis unit proposes an algorithm for analyzing the changes in soil and water loss patches based on time series classification. First, the algorithm separates the trend term and seasonal term in the time series to analyze the long-term change law of soil and water loss, providing a solid theoretical basis for change point detection. In soil and water loss monitoring, time series data is often affected by long-term trends and seasonal changes. Directly using the original data for change analysis is prone to noise interference. By constructing a trend model, the algorithm can identify the long-term change trend and dynamically adjust the threshold of change point detection in combination with land use information, making the detection results more in line with the actual soil and water loss situation. At the same time, the seasonal model suppresses seasonal noise unrelated to soil and water loss through Fourier decomposition to ensure the reliability of the analysis results. Second, the algorithm identifies the mutation points in the time series and divides the time series into stable subsequences to provide independent analysis units for subsequent classification. In this process, residual analysis is used to calculate the deviation between the observed value and the model prediction value at each time point, and the Bayesian Information Criterion (BIC) is combined to evaluate the rationality of different change point hypotheses, finally determining the optimal number and location of change points. Through this method, the system can more accurately determine the dynamic evolution of soil and water loss patches, distinguish long-term stable areas from areas vulnerable to erosion, and thus provide refined data support for soil and water conservation management. Second, the algorithm further improves the spatial accuracy of soil and water loss patches by combining pixel unmixing and patch classification techniques. The algorithm uses pixel unmixing technology to decompose the spectral characteristics of mixed pixels and extract the abundance information of different land cover classes, making the composition of each pixel clearer and more distinguishable. On this basis, the Dynamic Time Warping (DTW) classification method is used to measure the similarity between the time series to be classified and the reference time series and perform accurate classification. DTW classification can non-linearly align time series. Even if the data has time offsets and morphological changes, it can still effectively identify the change patterns of soil and water loss. Through this method, the system can accurately distinguish soil and water loss patches from background ground objects and significantly improve the accuracy of classification boundaries. In addition, the algorithm also uses prior land use information to optimize the patch boundaries to improve the spatial consistency of the classification results. By constructing a spatial probability fusion model, the probability of soil and water loss obtained by DTW classification is weighted and fused with the prior probability based on historical land use data, thereby reducing the false detection rate and improving the recognition accuracy. Finally, the level set method is used to optimize the patch boundaries. Through curvature constraint and spatial probability fusion, the boundaries of soil and water loss patches are made more in line with the morphology of the actual erosion area. Overall, the algorithm for analyzing the changes in soil and water loss patches based on time series classification not only improves the accuracy in the time dimension of soil and water loss monitoring but also enhances the boundary consistency of patches by combining spatial optimization techniques, achieving the accurate determination of soil and water loss patches. Brief Description of the Drawings

[0048] Figure 1 is a schematic structural diagram of the present invention; Detailed implementation manners

[0049] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0050] Please refer to Figure 1 , the present invention provides a soil and water loss patch landing system based on weak constraints of land use information, including a land use information collection module, a factor evaluation and analysis module, a time series analysis and patch landing module, and a data integration and visualization module. It is characterized in that: the land use information collection module is used to collect and integrate land use related data to ensure accurate land use basic information for subsequent analysis; the factor evaluation and analysis module includes an erosion modulus calculation unit and a factor classification and grading unit. The erosion modulus calculation unit proposes an erosion modulus calculation algorithm based on data fusion and pixel unmixing to accurately calculate the soil erosion modulus of each patch, ensuring that the calculation result of the erosion intensity can truly reflect the soil and water loss degree of different plots. The factor classification and grading unit is used to classify and grade each influencing factor to ensure that the soil and water loss factors are accurately mapped to the corresponding patches; the time series analysis and patch landing module includes a soil and water loss change analysis unit and a soil and water loss patch generation unit. The soil and water loss change analysis unit proposes a soil and water loss patch change analysis algorithm based on time series classification to analyze the patch changes, ensuring accurate capture of the patch change trend. The soil and water loss patch generation unit is used to generate soil and water loss patches according to the analysis results to ensure the accuracy and clear boundary of the patch landing; the data integration and visualization module is used to integrate and visually display the processed data to ensure that users can intuitively obtain the soil and water loss analysis results and treatment suggestions.

[0051] Refer to Figure 1 , further, the land use information collection module collects and processes land use data, including land types, plot information, and land use data, to ensure accurate and efficient land use basic data for subsequent generation and analysis of soil and water loss patches.

[0052] Refer to Figure 1, Further, the factor evaluation and analysis module includes an erosion modulus calculation unit. The erosion modulus calculation unit proposes an erosion modulus calculation algorithm based on data fusion and pixel unmixing. By combining remote sensing data from different sources with land use information, pixel unmixing technology is used to improve data resolution, and data fusion methods are used to optimize the calculation accuracy of each factor, ensuring that the soil erosion modulus of each patch can be accurately evaluated and providing accurate basic data support for the effective treatment of soil and water loss.

[0053] Refer to Figure 1 , Further, the erosion modulus calculation algorithm based on data fusion and pixel unmixing is as follows: First, data fusion is used to eliminate the uncertainty of a single data source and improve the reliability of the algorithm input parameters. A pixel-level weighted data fusion algorithm is constructed to spatially align and fuse the band reflectance of multi-spectral remote sensing images with the elevation information of the digital elevation model (DEM). Land use vector data is introduced as a weak constraint condition, and the fusion weight is optimized through spatial overlay analysis. The specific formula is expressed as:

[0054]

[0055] Among them, F i,j represents the comprehensive feature value obtained after multi-source data fusion at the pixel position (i, j), w k represents the contribution degree of the k-th type of data source during the fusion process, D k,i,j represents the specific value of the k-th type of data source at the pixel position (i, j). The data sources include different band data of remote sensing images and DEM elevation information data. i represents the row index of the pixel in space, used to locate the vertical position of the pixel in the two-dimensional grid data, j represents the column index of the pixel in space, used to locate the horizontal position of the pixel in the two-dimensional grid data, n represents the total number of data sources, k represents the data source number index, used to distinguish different data sources, α represents the weak constraint adjustment factor used to adjust the contribution degree of land use information to the fusion result, and L i,j represents the constraint effect of the land use type on the fusion result at the pixel position (i, j). By weakening the unreasonable area through land use information, the fused data can retain spectral features and terrain features. Then, pixel unmixing technology is used to separate the vegetation, soil, and water body endmember ratios in the mixed pixels, accurately extract the vegetation coverage, and calculate the slope and slope length in combination with the fused DEM data to provide dynamic parameters for the erosion modulus model. The specific formula for vegetation coverage unmixing is expressed as:

[0056]

[0057] Among them, V i,j represents the vegetation coverage of the pixel (i, j), R b,i,jDenoted as the reflectance of band b at pixel (i, j), S b Denoted as the standard reflectance of the bare soil endmember at band b, V b Denoted as the standard reflectance of the vegetation endmember at band b, m denotes the total number of bands. The specific formula for improving the nine-point slope length calculation is expressed as:

[0058]

[0059] Among them, L new Denoted as the slope length at the pixel center point, h z Denoted as the elevation value of the z-th pixel around the pixel center, z denotes the index of the number of pixels around the pixel center, h center Denoted as the elevation value of the central pixel, Denoted as the horizontal distance between the z-th pixel and the central pixel, cos denotes the cosine function, θ center Denoted as the slope angle of the central pixel. The specific formula for slope calculation is expressed as:

[0060]

[0061] Among them, S new Denoted as the improved slope factor at the pixel center point, Denoted as the elevation gradient, calculated from the elevation difference between the central pixel and the surrounding pixels. arctan denotes the arctangent function. By constraining the endmember selection through land use types, the interference of bare land and building areas on the unmixing results is avoided. Secondly, based on the framework of the Universal Soil Loss Equation USLE model, combined with the dynamic parameters of data fusion and pixel unmixing, the per-pixel calculation of the erosion modulus is realized, and the model coefficients are optimized using land use information. The erosion modulus formula is defined, a dynamic adjustment factor is introduced, and the erosion inhibition factor is adjusted according to land use types. The specific calculation formula is expressed as:

[0062] E = β·R·K·L new ·S new ·(1 - V)·P land

[0063] Among them, E denotes the erosion modulus, β denotes the dynamic calibration coefficient, R denotes the rainfall erosivity factor, K denotes the soil erodibility factor, V denotes the vegetation coverage, P land Denoted as the land use inhibition factor, P land Directly associated with weak constraints on land use information, the spatial differential correction of the erosion modulus is realized by dynamically assigning vector plot attributes. Finally, the erosion modulus raster map is classified according to national standards and converted into vector polygons. The spatial continuity of the polygons is optimized in combination with the land use boundary to meet management requirements. The specific formula for the vector polygon optimization rule is expressed as:

[0064] If A patch <is less than T and L land <is equal to cultivated land, it is merged into the adjacent forest land

[0065] Among them, A patch represents the vector patch area, T represents the area threshold, and L land represents the dominant land use type of the patch. By constraining the vector generalization process through the land use boundary, it ensures that the erosion patches are consistent with the actual plot boundaries of farmland and forest land, improving the implementation of the results.

[0066] Refer to Figure 1 , further, the factor evaluation and analysis module includes a factor classification and grading unit. The factor classification and grading unit classifies and grades various factors, including rainfall erosion factors, soil erodibility factors, slope factors, slope length factors, vegetation cover factors, soil and water conservation engineering measure factors, and tillage measure factors, to ensure that the factors related to soil erosion are accurately mapped onto the plot patches for subsequent patch analysis and processing.

[0067] Refer to Figure 1 , further, the time series analysis and patch implementation module includes a soil and water loss change analysis unit. The soil and water loss change analysis unit proposes an algorithm for analyzing the changes in soil and water loss patches based on time series classification. By analyzing the time series data of soil and water loss patches over the years, it identifies the change patterns and trends of the patches, and combines with the land use change situation to ensure that it can accurately judge the dynamic changes and stability of soil and water loss patches, providing a reliable basis for the long-term monitoring and control of soil and water loss.

[0068] Refer to Figure 1 , further, the algorithm for analyzing the changes in soil and water loss patches based on time series classification is as follows: First, by separating the trend term and seasonal term in the time series, it analyzes the long-term change law of soil and water loss, providing a basis for change point detection. The specific formula is expressed as:

[0069]

[0070] Among them, T(t) represents the global trend of the time series, ε represents the basic intercept, η represents the linear slope, δ x represents the trend change amplitude of the x-th change point, t represents the time variable, X represents the total number of change points detected in the time series, τ x represents the time position of the x-th change point, representing the time point when the trend changes suddenly. I represents the indicator function. The I function is used to introduce a change in the trend after the change point. When the time t is greater than or equal to the change point τ x , the function value is 1, otherwise it is 0. The specific formula of the seasonal model is expressed as:

[0071]

[0072] Among them, S(t) represents the seasonal component of the time series at time t, which represents the mathematical expression of the periodic change in the time series. n represents the order of the Fourier series, which is used to describe the complexity of the seasonal change. N represents the highest order of the Fourier series. a n represents the coefficient of the nth-order cosine term, which is used to describe the amplitude of the cosine component in the seasonal change. b n represents the coefficient of the nth-order sine term, which is used to describe the amplitude of the sine component in the seasonal change. represents the nth-order cosine term, which is used to capture the cosine fluctuations in the seasonal change. represents the nth-order sine term, which is used to capture the sine fluctuations in the seasonal change. P represents the seasonal period, and the change point τ is adjusted according to the land use type through the trend model x The detection threshold ensures that the change point is related to soil and water loss. Through the seasonal model: adjust the Fourier coefficients a n and b n to suppress the seasonal noise unrelated to soil and water loss. Then, identify the mutation points of soil and water loss in the time series and segment them into stable subsequences to provide independent analysis units for classification. The specific formula is expressed as:

[0073] R(t) = V fused (t) - [T(t) + S(t)]

[0074] Among them, R(t) represents the residual value at time t, which represents the difference between the observed value and the model prediction value, and is expressed as the pixel value after multi-source data fusion at time t. T(t) represents the trend term at time t, which represents the long-term change trend of the time series. S(t) represents the seasonal term at time t, which represents the periodic change in the time series. The specific formula for change point determination is expressed as:

[0075] BIC = -2ln(L) + ζln(T)

[0076] Among them, BIC represents the Bayesian information criterion, which is a statistic used to evaluate the goodness of fit of the model. ln(·) represents the logarithmic function. L represents the likelihood function, which represents the fitting degree of the algorithm to the observed data. ζ represents the number of model parameters, and T represents the total length of the time series.

[0077] Calculate the difference between the observed value and the model predicted value at each time point through residual analysis, identify outliers and change points, perform change point determination calculation based on the results of residual analysis, evaluate the rationality of different change point hypotheses through the Bayesian Information Criterion (BIC), and select the optimal number and location of change points. ln(·) represents the natural logarithm function. Secondly, solve the mixed pixel problem through pixel unmixing and patch classification, accurately distinguish soil erosion patches from background ground objects, and improve the accuracy of classification boundaries. The specific formula for linear unmixing is as follows:

[0078]

[0079] Among them, V mixed (t) represents the mixed pixel value at time t, f c represents the abundance of the c-th type of ground object, which represents the proportion of the c-th type of ground object in the mixed pixel. V c (t) represents the pure spectral time series value of the c-th type of ground object at time t, ε represents the noise term, which represents the residual part that the algorithm fails to explain, C represents the total number of ground object categories, and the specific formula for dynamic time warping classification is as follows:

[0080]

[0081] Among them, DWB(O,R) represents the dynamic time warping distance between time series O and R, O represents the time series to be classified, R represents the reference time series, w represents the alignment path, which represents the optimal alignment method between time series O and R, U represents the total length of the alignment path, u represents the length index of the alignment path, represents the value of O corresponding to the u-th point in the alignment path w, represents the value of R corresponding to the u-th point in the alignment path w, represents minimizing the total distance of the alignment path w. By decomposing the spectral characteristics of the mixed pixel, the abundance of each ground object type is obtained, and the mixed pixel problem is solved. By non-linearly aligning the time series, the similarity between the time series to be classified and the reference series is measured, and high-precision classification is achieved. Finally, fuse the prior information of land use, optimize the patch boundary, and generate high-precision soil erosion landing patches. The specific formula for spatial probability fusion is as follows:

[0082] P final (x,y) = λ·P class (x,y) + (1 - λ)·P landuse (x,y)

[0083] Among them, P final (x,y) represents the final probability value at position (x,y), which represents the probability that the position (x,y) belongs to the soil erosion patch. P class(x, y) represents the probability of soil and water loss obtained by dynamic time warping classification at the position (x, y), P landuse (x, y) represents the prior probability at the position (x, y) based on historical land use data, λ represents the fusion weight, and the specific formula for level set boundary optimization is expressed as:

[0084]

[0085] where φ t (x, y) represents the level set function value at the position (x, y) at time t, φ t+1 (x, y) represents the level set function value at the position (x, y) at time t + 1, η represents the evolution rate parameter, which controls the step size of the level set function update, represents the curvature term of the level set function, representing the change in boundary curvature, represents the gradient of the level set function, representing the rate of change of the function value in space, represents the magnitude of the gradient of the level set function, κ represents the weight parameter, which is used to control the influence of the probability term on the boundary evolution, P final represents the final probability value at the position (x, y). By data fusion, the spatio-temporal continuity is enhanced. Pixel unmixing solves the problem of mixed pixels. Using weak land use constraints, false detections are suppressed through probability fusion to ensure that the patch boundaries are consistent with the field. Combining temporal analysis and spatial optimization, the accurate landing of soil and water loss patches is achieved.

[0086] Refer to Figure 1 , further, the temporal analysis and patch landing module includes a soil and water loss patch generation unit. The soil and water loss patch generation unit generates accurate soil and water loss patches by combining different soil and water loss factors with land use information according to the analysis results of the soil and water loss change analysis unit, and performs smoothing and merging processing on the patches to ensure that the generated patch boundaries are clear and conform to the actual distribution of soil and water loss.

[0087] Refer to Figure 1 , further, the data integration and visualization module integrates all the generated data and analysis results for visual display to ensure that users can intuitively view the soil and water loss status, erosion intensity, and patch changes, thereby providing effective decision support to help implement accurate soil and water loss control measures.

[0088] In specific use, first, the land use information collection module is used to collect and integrate land use-related data to ensure accurate land use basic information for subsequent analysis. Then, the factor evaluation and analysis module includes an erosion modulus calculation unit and a factor classification and grading unit. The erosion modulus calculation unit proposes an erosion modulus calculation algorithm based on data fusion and pixel unmixing to accurately calculate the soil erosion modulus of each patch, ensuring that the calculation result of the erosion intensity can truly reflect the soil and water loss degree of different plots. The factor classification and grading unit is used to classify and grade each influencing factor to ensure accurate mapping of the soil and water loss factors to the corresponding patches. Secondly, the time series analysis and patch landing module includes a soil and water loss change analysis unit and a soil and water loss patch generation unit. The soil and water loss change analysis unit proposes a soil and water loss patch change analysis algorithm based on time series classification to analyze the patch changes, ensuring accurate capture of the patch change trend. The soil and water loss patch generation unit is used to generate soil and water loss patches according to the analysis results to ensure the accuracy and clear boundary of patch landing. Finally, the data integration and visualization module is used to integrate and visually display the processed data to ensure that users can intuitively obtain the soil and water loss analysis results and treatment suggestions.

[0089] Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A soil erosion patch landing system based on weak constraints of land use information, comprising a land use information collection module, a factor evaluation and analysis module, a time series analysis and patch landing module, and a data integration and visualization module, characterized in that: The land use information collection module is used to collect and integrate land use-related data to ensure accurate land use basic information for subsequent analysis; the factor evaluation and analysis module includes an erosion modulus calculation unit and a factor classification and grading unit. The erosion modulus calculation unit proposes an erosion modulus calculation algorithm based on data fusion and pixel unmixing, which is used to accurately calculate the soil erosion modulus of each patch, ensuring that the calculation results of the erosion intensity can truly reflect the soil and water loss degree of different plots. The factor classification and grading unit is used to classify and grade each influencing factor to ensure that the soil and water loss factors are accurately mapped to the corresponding patches; the time series analysis and patch landing module includes a soil and water loss change analysis unit and a soil and water loss patch generation unit. The soil and water loss change analysis unit proposes a soil and water loss patch change analysis algorithm based on time series classification to analyze the patch changes, ensuring the accurate capture of the patch change trend. The soil and water loss patch generation unit is used to generate soil and water loss patches according to the analysis results to ensure the accuracy of patch landing and clear boundaries; the data integration and visualization module is used to integrate and visually display the processed data to ensure that users can intuitively obtain the soil and water loss analysis results and treatment suggestions.

2. The soil erosion patch landing system based on weak constraints of land use information according to claim 1, wherein: The land use information collection module ensures accurate and efficient land use basic data for subsequent soil and water loss patch generation and analysis by collecting and processing land use data, including land types, plot information, and land use data.

3. The soil erosion patch landing system based on weak constraints of land use information according to claim 1, characterized in that: The factor evaluation and analysis module includes an erosion modulus calculation unit and a factor classification and grading unit. The erosion modulus calculation unit proposes an erosion modulus calculation algorithm based on data fusion and pixel unmixing. By combining remote sensing data from different sources with land use information, pixel unmixing technology is used to improve data resolution, and data fusion methods are used to optimize the calculation accuracy of each factor, ensuring that the soil erosion modulus of each patch can be accurately evaluated and providing accurate basic data support for the effective treatment of soil and water loss; the factor classification and grading unit classifies and grades various factors, including rainfall erosion factors, soil erodibility factors, slope factors, slope length factors, vegetation cover factors, soil and water conservation engineering measure factors, and tillage measure factors, to ensure that the factors related to soil erosion are accurately mapped to the plot patches for subsequent patch analysis and processing.

4. The soil erosion patch landing system based on weak constraints of land use information according to claim 3, characterized in that, First, data fusion is used to eliminate the uncertainty of a single data source, improve the reliability of the algorithm input parameters, construct a pixel-level weighted data fusion algorithm, spatially align and fuse the band reflectance of the multispectral remote sensing image with the elevation information of the digital elevation model (DEM), introduce land use vector data as a weak constraint condition, and optimize the fusion weight through spatial overlay analysis. The specific formula is expressed as: Among them, F i,j It is represented as the comprehensive feature value obtained after multi-source data fusion at the pixel position (i, j), w k It is expressed as the contribution of the k-th data source in the fusion process, D k,i,j It is represented as the specific value of the k-th data source at the pixel position (i, j). The data sources include different bands of remote sensing images and DEM elevation information data. i is represented as the row index of the pixel in space, which is used to locate the vertical position of the pixel in the two-dimensional raster data. j is represented as the column index of the pixel in space, which is used to locate the horizontal position of the pixel in the two-dimensional raster data. n is represented as the total number of data sources. k is represented as the number index of data sources, which is used to distinguish different data sources. α is represented as a weak constraint adjustment factor used to adjust the contribution of land use information to the fusion result. L i,j It is expressed as the constraint effect of land use type on the fusion result at the pixel position (i, j). The unreasonable area is weakened by land use information. The fused data can retain the spectral characteristics and terrain characteristics. Then, the pixel unmixing technology is used to separate the proportion of vegetation, soil, and water end members in the mixed pixels, accurately extract vegetation coverage, and calculate the slope and slope length in combination with the fused DEM data to provide dynamic parameters for the erosion modulus model. The specific formula for vegetation coverage unmixing is expressed as: Among them, V i,j represents the vegetation coverage of pixel (i, j), R b,i,j represents the reflectance of band b at pixel (i, j), S b represents the standard reflectance of the soil endmember in band b, V b represents the standard reflectance of the vegetation endmember in band b, m represents the total number of bands, and the specific formula for improving the nine-point slope length calculation is expressed as: Among them, L new represents the slope length of the pixel center point, h z represents the elevation value of the z-th pixel around the pixel center, z represents the index of the number of pixels around the pixel center, h center represents the elevation value of the central pixel, represents the horizontal distance between the z-th pixel and the central pixel, cos represents the cosine function, θ center represents the slope angle of the central pixel, and the specific formula for slope calculation is as follows: Among them, S new represents the improved slope factor of the pixel center point, represents the elevation gradient, which is calculated through the elevation difference between the central pixel and the surrounding pixels. Arctan represents the arctangent function. By constraining the endmember selection based on the land use type, the interference of bare land and building areas on the unmixing result is avoided. Secondly, based on the framework of the Universal Soil Loss Equation (USLE) model, combined with the dynamic parameters of data fusion and pixel unmixing, the per-pixel calculation of the erosion modulus is realized, and the model coefficients are optimized using land use information. The erosion modulus formula is defined, a dynamic adjustment factor is introduced, and the erosion inhibition factor is adjusted according to the land use type. The specific calculation formula is expressed as: E = β·R·K·L new ·S new ·(1 - V)·P land Among them, E represents the erosion modulus, β represents the dynamic calibration coefficient, R represents the rainfall erosivity factor, K represents the soil erodibility factor, V represents the vegetation coverage, and P land represents the land use inhibition factor, and P land is directly related to the weak constraint of land use information. Through the dynamic assignment of vector plot attributes, the spatial differential correction of the erosion modulus is realized. Finally, the erosion modulus raster map is classified according to national standards and converted into vector patches. Combining with the land use boundary, the spatial continuity of the patches is optimized to meet the management requirements. The specific formula of the vector patch optimization rule is expressed as: If A patch <T and L land = cultivated land, then merge it into the adjacent forest land Among them, A patch represents the vector patch area, T represents the area threshold, and L land represents the dominant land use type of the patch. By constraining the vector generalization process with the land use boundary, it ensures that the erosion patches are consistent with the actual plot boundaries of farmland and forest land, and improves the implementation of the results.

5. A soil erosion patch landing system based on weak constraints of land use information according to claim 1, characterized in that: The above-mentioned time series analysis and mapping module for soil erosion patches includes a soil erosion change analysis unit and a soil erosion patch generation unit. The soil erosion change analysis unit proposes an algorithm for analyzing the changes in soil erosion patches based on time series classification. By analyzing the time series data of soil erosion patches over the years, it identifies the change patterns and trends of the patches, and combines the land use changes to ensure that the dynamic changes and stability of the soil erosion patches can be accurately judged, providing a reliable basis for the long-term monitoring and treatment of soil erosion. The soil erosion patch generation unit generates accurate soil erosion patches by combining different soil erosion factors with land use information according to the analysis results of the soil erosion change analysis unit, and performs smoothing and merging processing on the patches to ensure that the generated patch boundaries are clear and conform to the actual distribution of soil erosion.

6. The soil erosion patch landing system based on weak constraints of land use information according to claim 5, characterized in that, First, by separating the trend term and seasonal term in the time series, the long-term change law of soil erosion is analyzed, providing a basis for change point detection. The specific formula is expressed as: Among them, T(t) represents the global trend of the time series, ε represents the basic intercept, η represents the linear slope, and δ x represents the magnitude of the trend change at the x-th change point, t represents the time variable, X represents the total number of change points detected in the time series, and τ x represents the time position of the x-th change point, which represents the time point when the trend undergoes a sudden change. I represents the indicator function. The I function is used to introduce a change in the trend after the change point. When the time t is greater than or equal to the change point τ x , the function value is 1, otherwise it is 0. The specific formula of the seasonal model is expressed as: Among them, S(t) represents the seasonal component of the time series at time t, which is a mathematical expression of the periodic changes in the time series. n represents the order of the Fourier series, which is used to describe the complexity of the seasonal changes. N represents the highest order of the Fourier series. a n represents the coefficient of the nth-order cosine term, which is used to describe the amplitude of the cosine component in the seasonal changes. b n represents the coefficient of the nth-order sine term, which is used to describe the amplitude of the sine component in the seasonal changes. represents the nth-order cosine term, which is used to capture the cosine fluctuations in the seasonal changes. represents the nth-order sine term, which is used to capture the sine fluctuations in the seasonal changes. P represents the seasonal period, and the change point τ is adjusted according to the land use type through the trend model x of the detection threshold to ensure that the change point is related to soil and water loss. Through the seasonal model: adjust the Fourier coefficients a n and b n , suppress the seasonal noise unrelated to soil and water loss. Then, identify the mutation points of soil and water loss in the time series, segment them into stable subsequences, and provide independent analysis units for classification. The specific formula is expressed as: R(t) = V fused (t) - [T(t) + S(t)] Among them, R(t) represents the residual value at time t, representing the difference between the observed value and the model prediction value, which is expressed as the pixel value after multi-source data fusion at time t. T(t) represents the trend term at time t, representing the long-term change trend of the time series. S(t) represents the seasonal term at time t, representing the periodic change in the time series. The specific formula for change point determination is expressed as: BIC = -2ln(L) + ζln(T) Among them, BIC represents the Bayesian information criterion, a statistic for evaluating the goodness of fit of the model. ln(·) represents the logarithmic function. L represents the likelihood function, representing the fitting degree of the algorithm to the observed data. ζ represents the number of model parameters. T represents the total length of the time series. By calculating the difference between the observed value and the model prediction value at each time point through residual analysis, abnormal points and change points are identified. Through change point determination calculation based on the results of residual analysis, the rationality of different change point hypotheses is evaluated by the Bayesian information criterion BIC, and the optimal number and location of change points are selected. ln(·) represents the natural logarithmic function. Secondly, through pixel unmixing and patch classification, the problem of mixed pixels is solved, the soil erosion patches and background ground objects are accurately distinguished, and the accuracy of the classification boundary is improved. The specific formula for linear unmixing is expressed as: Among them, V mixed (t) represents the mixed pixel value at time t, and f c represents the abundance of the c-th type of ground object, representing the proportion of the c-th type of ground object in the mixed pixel. V c (t) represents the pure spectral time series value of the c-th type of ground object at time t. ∈ represents the noise term, representing the residual part that the algorithm fails to explain. C represents the total number of ground object categories. The specific formula for dynamic time warping classification is expressed as: Among them, DWB(O, R) represents the dynamic time warping distance between time series O and R, O represents the time series to be classified, R represents the reference time series, w represents the alignment path, which represents the optimal alignment method between time series O and R, U represents the total length of the alignment path, and u represents the length index of the alignment path. represents the value of O corresponding to the u-th point in the alignment path w. represents the value of R corresponding to the u-th point in the alignment path w. represents minimizing the total distance of the alignment path w. By decomposing the spectral features of mixed pixels, the abundances of various ground object types are obtained to solve the mixed pixel problem. By non-linearly aligning time series, the similarity between the time series to be classified and the reference series is measured to achieve high-precision classification. Finally, by fusing the prior information of land use, the patch boundaries are optimized to generate high-precision soil erosion patch maps. The specific formula for spatial probability fusion is as follows: P final (x,y) = λ·P class (x,y) + (1 - λ)·P landuse (x,y) Among them, P final (x, y) represents the final probability value at the position (x, y), representing the probability that the position (x, y) belongs to the soil erosion patch. P class (x, y) represents the soil erosion probability obtained by dynamic time warping classification at the position (x, y). P landuse (x, y) represents the prior probability based on historical land use data at the position (x, y). λ represents the fusion weight. The specific formula for level set boundary optimization is expressed as: Among them, φ t (x, y) represents the value of the level set function at the position (x, y) at time t, and φ t+1 (x, y) represents the value of the level set function at the position (x, y) at time t + 1. η represents the evolution rate parameter, which controls the step size of the update of the level set function. represents the curvature term of the level set function, representing the change in the boundary curvature. represents the gradient of the level set function, representing the rate of change of the function value in space. represents the magnitude of the gradient of the level set function. κ represents the weight parameter, which is used to control the influence of the probability term on the boundary evolution. P final represents the final probability value at the position (x, y). By data fusion, the spatio-temporal continuity is enhanced. Pixel unmixing solves the problem of mixed pixels. The weak land use constraint is used to suppress false detections through probability fusion, ensuring that the patch boundary is consistent with the field situation. Combining temporal sequence analysis and spatial optimization, the accurate location of the soil erosion patches is achieved.

7. A soil erosion patch landing system based on weak constraints of land use information according to claim 1, characterized in that The data integration and visualization module integrates all the generated data and analysis results for visual display, ensuring that users can intuitively view the soil erosion status, erosion intensity, and patch changes, thereby providing effective decision support and helping to implement precise soil erosion control measures.

Images