Salinization space analysis method and system based on GIS (Geographic Information System)

Abstract

The invention relates to the technical field of image analysis, in particular to a GIS-based salinization spatial analysis method and system, and the method comprises the following steps: extracting a pattern spot spectrum based on a remote sensing image, comparing the pattern spot spectrum with salinity attributes, recognizing an abnormal pattern spot, analyzing the spectrum index distribution of the abnormal pattern spot, setting a correction factor, adjusting the spectrum through a correction vector, and progressively increasing the weight of a sensitive wave band; according to the method, the relation between the spectrum and the salinity is enhanced through combined recognition of the spectral reflectivity sequence and the salinity attribute dispersion, the salinity characteristics are enhanced through sensitive band weight increasing and error gradient quantification, and the salinity risk partition layer is obtained through the combination recognition of the spectral reflectivity sequence and the salinity attribute dispersion and the combination recognition of the spectrum reflectivity sequence and the salinity attribute dispersion. Spatial interpretation force is enhanced through weighted fusion of ion components and environmental factors, and risk division consistency is ensured through coupling of action grades and salinity evaluation values, so that salinization recognition accuracy and feature definition are improved, and spatial partition stability and research and judgment effectiveness are optimized.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and in particular to a GIS-based method and system for spatial analysis of salinization. Background Technology

[0002] Image analysis technology primarily involves processing, understanding, and interpreting digital images using computer algorithms to extract useful information and support decision-making. Its core aspects include image acquisition, preprocessing, feature extraction, segmentation, classification, and pattern recognition. In practical applications, this field first uses sensors or cameras to acquire raw image data. Then, preprocessing steps such as filtering, enhancement, and correction remove noise and distortion. Next, features such as edges, textures, colors, and shapes are extracted. Image regions are segmented to distinguish different objects. Finally, statistical methods or machine learning algorithms are used to achieve object recognition and scene understanding, widely applied in remote sensing monitoring, medical diagnosis, and agricultural assessment. Traditional salinization spatial analysis methods refer to the technology of spatially assessing the distribution and extent of saline-alkali farmland using geographic information systems. This method, aimed at monitoring and early warning of farmland salinization, first acquires remote sensing image data via satellite or drones. Then, radiometric calibration and atmospheric correction are performed on the images to obtain accurate ground reflectance. The normalized vegetation index and soil salinity index are then calculated to identify saline areas. Finally, the indices are overlaid with soil sampling point data and topographic layers, and a salinization distribution map is generated through Kriging interpolation or buffer analysis.

[0003] Existing technologies rely on the static correspondence between spectral indices and soil samples for salinization analysis. However, fluctuations in spectral reflectance over time are often out of sync with changes in salt ions, leading to a disconnect between spectral response and salinity properties. The offset of index distribution to anomalous patches fails to reflect the true discrete characteristics of salinity components. The lack of specificity in spectral correction results in insufficient accuracy in identifying salinized areas. The index superposition interpolation process does not consider the differences in contributions from sensitive bands and the sources of reconstruction errors. Over-reliance on topography and sampling point distribution in spatial inference leads to a simplistic expression of salinization causes. The causal weights fail to reflect the impact of differences in ion combinations on salinization formation mechanisms. The determination of zoning thresholds based solely on index changes can easily lead to ambiguity in risk category classification, affecting the reliability and spatial indicative significance of salinization assessment. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide a GIS-based spatial analysis method for salinization, comprising the following steps: To achieve the above objectives, the present invention adopts the following technical solution: a GIS-based spatial analysis method for salinization, comprising the following steps: S1: Based on remote sensing image patch units, extract the spectral reflectance sequence of patches within the current monitoring period, identify the measured total soil salinity, content of eight major ion components and pH value of the corresponding patch location, compare spectral fluctuations with salt separation divergence, identify mismatch areas, and obtain a set of abnormal salinity response patches. S2: Based on the set of anomalous patches in salinity response, analyze the high-throughput patch index, compare the offset direction of the normalized vegetation index and the soil salinity index, set correction factors for patches that deviate from the upper and lower boundaries of the main distribution cluster, and obtain the patch spectral correction vector. S3: Based on the spectral correction vector of the patch, the original spectrum is calculated, and the patch is determined to be in a weakened state when the corrected spectrum is lower than the preset threshold. The salt sensitivity weight is increased by step size, and the reconstruction error and salt fitting residual after adjustment are recorded to obtain the salt feature enhancement gradient distribution dataset. S4: Based on the enhanced gradient distribution dataset of salinity features, combined with ground-measured ion data, the geographical location, parent material, irrigation water source and salt ion combination are weighted and fused. According to the contribution of factors to the salinity formation mechanism, the level of influence is assigned to obtain the weighted structure system for salinity causation analysis.

[0005] As a further embodiment of the present invention, the abnormal salinity response patch set includes abnormal patch number, spatial clustering range, and mismatch type; the patch spectral correction vector includes correction factor value, adjustment band identifier, and patch clustering label; the salinity feature enhancement gradient distribution dataset includes gradient direction, gradient amplitude range, and response enhancement level; and the weighted structure system for salinity causation analysis includes ion combination weight, geographical factor level, and water source influence order.

[0006] As a further aspect of the present invention, the step of obtaining the salt-response anomaly patch set specifically includes: S101: Based on remote sensing image patch units, extract the time series of reflectance of patches in the visible to shortwave infrared bands within the current monitoring period, and simultaneously match the measured records of total soil salinity, eight major ion components and pH value at the same patch location to identify the multidimensional attribute vector of each patch and obtain the patch attribute matrix. S102: Based on the attribute matrix of the patch, calculate the ratio of the standard deviation of the spectral reflectance of the patch to the coefficient of variation of the corresponding salinity attribute. Patches with a ratio exceeding a preset threshold are marked as spectral and salinity mismatch units, and the remaining patches are divided into stable response units to generate patch response consistency labels. S103: Based on the consistency label of the patch response, extract the spatial coordinates and attribute features of the mismatched units, aggregate them according to the adjacency relationship to form a continuous region, label the mismatch dominant factor type, and obtain the salt response abnormal patch set.

[0007] As a further aspect of the present invention, the step of obtaining the spectral correction vector of the patch is specifically as follows: S201: Based on the set of anomalous patches in response to salinity, the joint distribution density of normalized vegetation index and soil salinity index of all patch units is statistically analyzed, the boundary contour of the main cluster of index distribution is identified, and the patches located outside the contour are uniquely encoded and bound to the patches to form an index offset identifier set. S202: Based on the index offset identifier set, determine the azimuth angle of each offset patch in the index plane. If the azimuth angle points to the high vegetation and low salinity quadrant, it is marked as vegetation interference type. If it points to the low vegetation and high salinity quadrant, it is marked as bare soil enhancement type, thus obtaining the patch interference type classification. S203: Based on the classification of the interference types of the patches, reduce the weight of the near-infrared band and increase the gain of the short-wave infrared band for vegetation interference patches, and reverse the band weight for bare soil enhancement patches. Record the band adjustment amount and direction, and integrate them into a patch spectral correction vector.

[0008] As a further aspect of the present invention, the steps for obtaining the salt feature enhancement gradient distribution dataset are as follows: S301: Based on the spectral correction vector of the patch, perform band weight substitution calculation on each patch, identify the corrected spectral curve, screen patches whose integral performance in the salt-sensitive region of the near-infrared front segment and the mid-infrared front segment is lower than the benchmark level, and establish a list of feature-weakened patches. S302: Based on the list of weakened feature patches, for each patch, the weight coefficient of the sensitive band is adjusted step by step with a preset threshold, the spectral reconstruction error and the salt regression residual based on the eight major ion components are calculated, and the error sequence is recorded in the order of adjustment to form error and weight trajectory data. S303: Based on the error and weight trajectory data, calculate the ratio of the decrease in residual to the increase in error between adjacent weight steps. When the ratio exceeds a preset threshold, retain the corresponding weight configuration and combine the map features that meet the conditions and their weight adjustment ranges to obtain the salt feature enhancement gradient distribution dataset.

[0009] As a further aspect of the present invention, the steps for obtaining the weighted structural system for salt formation analysis are specifically as follows: S401: Based on the salinity feature enhancement gradient distribution dataset, spatially register the spectral correction vector of the patch with the ground-measured molar concentration vectors of the eight ions, and superimpose the administrative division to which the patch belongs, the parent material lithology category, and the irrigation canal system number to construct a four-dimensional genetic element record. S402: Based on the four-dimensional causal element record, the co-occurrence frequency of bicarbonate and sodium ions, the antagonistic ratio of sulfate and calcium and magnesium ions, and the correlation slope between chloride ions and conductivity are statistically analyzed. The dominant combination type is divided according to the intensity of ion interaction, and the dominant combination type and intervention level are obtained. S403: Based on the dominant combination type and intervention level, extract the basic weights of eight ion distributions, set the soil parent material distribution adjustment weights and irrigation water source distribution correction weights, link and integrate the three types of weights and determine the level of action to obtain the weighted structure system for salinization cause analysis.

[0010] As a further aspect of the present invention, the co-occurrence frequency of bicarbonate and sodium ions refers to the frequency at which bicarbonate and sodium ions appear simultaneously within the target spatial range by comparing the spectral data of the map patch with the ground-based measured data. The antagonistic ratio between sulfate and calcium and magnesium ions refers to the ratio of their antagonism in the patch obtained by analyzing the ion concentration data of the patch and analyzing the interaction strength between sulfate and calcium and magnesium ions. The slope of the correlation between chloride ions and conductivity refers to the slope between chloride ion concentration and conductivity obtained by regression analysis of the relationship between chloride ion concentration and conductivity.

[0011] As a further aspect of the present invention, the method further includes step S5: S5: Based on the weighted structure system of the salinity analysis, the salinity evaluation value of the map patch is multiplied by the level of action. The product result is then assigned to the threshold range of mild, moderate and severe salinization. The map patch risk categories are divided according to the determination results to obtain the salinization risk zoning layer. The salinization risk zoning layer includes risk category code, threshold attribution range, and patch spatial index.

[0012] As a further aspect of the present invention, the steps for obtaining the salinization risk zoning layer are specifically as follows: S501: Based on the weighted structure system for the analysis of salinity formation, calculate the total salt equivalent of eight ions for each patch, generate a salinity index by combining the pH correction term, and compare the salinity index with the preset thresholds of mild, moderate and severe salinity to obtain the interval assignment result. S502: Based on the interval attribution results, assign a corresponding risk category code to each map patch, merge similar map patches according to spatial adjacency, and obtain a salinization risk zoning layer.

[0013] A GIS-based spatial analysis system for salinization includes: The spectral and salinity mismatch identification module is based on remote sensing image patch units. It extracts the spectral reflectance sequence of the patches within the current monitoring period, obtains the total soil salinity, eight major ion components and pH value of the corresponding patch location, and identifies the mismatch area between the spectral and salinity relationship by comparing the fluctuation amplitude of the patch spectral response with the dispersion of salinity attributes, thus obtaining a set of abnormal salinity response patches. The spectral correction factor generation module analyzes the joint distribution of normalized vegetation index and soil salinity index of the patches in the high-throughput remote sensing patch unit based on the set of salinity response anomalies. It compares the offset direction of the boundary of the main distribution cluster corresponding to the index points, reduces the weight of vegetation sensitive bands for patches with upper boundary offset, and records the adjustment content for patches with lower boundary offset, thus obtaining the spectral correction vector of the patch. Based on the spectral correction vector of the patch, the salt feature gradient calculation module performs band weight substitution operation on the patch, filters the patches whose salt-sensitive band response is lower than the threshold after correction, increases the weight by a fixed step size and records the spectral reconstruction error and salt fitting residual, calculates the error descent gradient, and obtains the salt feature enhancement gradient distribution dataset. The causal weight fusion module, based on the salinity feature enhancement gradient distribution dataset, spatially registers the patch spectral correction vector with the measured data of eight ions, overlays geographical location, soil parent material type and irrigation water source type, and obtains a weighted structure system for salinity causal analysis based on ion co-occurrence frequency and geographical intervention level. The risk zoning generation module calculates the salinity index of the map patches based on the weighted structure system of the salinity cause analysis, classifies the index with the threshold ranges of mild, moderate and severe, assigns risk category codes according to the judgment results, merges adjacent map patches, and outputs the salinization risk zoning layer.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the consistency between the spectrum and salinity is enhanced by jointly identifying the spectral reflectance sequence and the dispersion of salinity attributes. The reliability of the spectral response is improved by establishing a differentiated correction vector through the analysis of the spectral index shift direction of abnormal patches. The enhancement of salinity characteristics by spectral reconstruction is achieved by increasing the weight of sensitive bands and quantifying the error gradient. The spatial explanatory power is enhanced by forming a more causal expressive structure system through the weighted fusion of ionic component data and environmental factors. The consistency between the level of action and the salinity evaluation value is ensured by ensuring that the risk category classification is consistent with the salinization process. This improves the clarity of salinity characteristic expression, enhances the accuracy of salinization identification, optimizes the stability and distinguishability of spatial partitioning results, and improves the pertinence and effectiveness of salinity risk assessment. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] Please see Figure 1 This invention provides a GIS-based spatial analysis method for salinization, comprising the following steps: S1: Based on remote sensing image patch units, extract the spectral reflectance sequence of patches within the current monitoring period, identify the measured total soil salinity, content of eight major ion components and pH value of the corresponding patch location, and identify the mismatch area between spectrum and salinity by comparing the fluctuation amplitude of patch spectral response and the dispersion of salinity attributes, thus obtaining a set of abnormal salinity response patches. S2: Based on the salt response anomaly patch set, analyze the spectral index distribution pattern of patches in high-throughput remote sensing patch units, compare the distribution offset direction of normalized vegetation index and soil salinity index in the patch set, and set spectral correction factors for patches whose indices deviate from the upper and lower bounds of the main distribution cluster to obtain the patch spectral correction vector. S3: Based on the spectral correction vector of the patch, perform algebraic operations on the original spectral reflectance of the patch band by band. For patches whose spectral response after correction is lower than the preset threshold, determine their salt feature weakening state. Increase the weight of the salt sensitive band by a fixed step size, record the spectral reconstruction error and salt fitting residual after each weight adjustment, calculate the error descent gradient caused by the weight change, and obtain the salt feature enhancement gradient distribution dataset. S4: Based on the enhanced gradient distribution dataset of salinity features, combined with the spectral correction vector of the patch and the ground-measured data of eight major ion components, the geographical location of the patch, the type of parent material, the type of irrigation water source and the combination of salt ions are weighted and fused. The role level is assigned according to the contribution order of factors in the salinity formation mechanism, and the weighted structure system of salinity causation analysis is obtained. S5: Based on the weighted structure system of salinization cause analysis, the salinity evaluation value of the map patch is multiplied by the effect level, and the product result is assigned to the threshold range of mild, moderate and severe salinization. The map patch risk category is divided according to the judgment result to obtain the salinization risk zoning layer.

[0023] The salinity response anomaly patch set includes anomaly patch number, spatial clustering range, and mismatch type; the patch spectral correction vector includes correction factor value, adjustment band identifier, and patch clustering label; the salinity feature enhancement gradient distribution dataset includes gradient direction, gradient amplitude range, and response enhancement level; the salinity causation analysis weighted structure system includes ion combination weight, geographical factor level, and water source influence order; and the salinization risk zoning layer includes risk category code, threshold attribution range, and patch spatial index.

[0024] Please see Figure 2 The specific steps for obtaining the salt-response anomaly patch set are as follows: S101: Based on remote sensing image patch units, extract the time series of reflectance of patches in the visible to shortwave infrared bands within the current monitoring period, and simultaneously match the measured records of total soil salinity, eight major ion components and pH value at the same patch location to identify the multidimensional attribute vector of each patch and obtain the patch attribute matrix. Based on remote sensing image patch units, multispectral remote sensing image data from Sentinel-2 and Landsat-8 satellites within the monitoring area were retrieved. Radiometric calibration and atmospheric correction were performed on the image data to eliminate interference from atmospheric scattering and aerosols on ground reflectance. The images were resampled to a spatial resolution of 10 meters and divided into independent pixel grid units. For each grid unit, reflectance values ​​of the visible light band, near-infrared band, and short-wave infrared band were extracted and stacked in the time dimension to construct a high-dimensional spectral feature vector. Precise three-dimensional coordinates of ground soil sample points were acquired using GPS RTK technology, with a planar position matching tolerance of 0.5 meters. The measured total soil salinity and pH values ​​were then compared with those of the ground. H-values ​​and chemical analysis data of eight major ions, including carbonate, bicarbonate, chloride, sulfate, calcium, magnesium, potassium, and sodium ions, are mapped to corresponding image pixel grids. For grid cells lacking measured data, a co-kriging interpolation algorithm is used to spatially estimate the attribute values ​​of surrounding effective sampling points, ensuring that each grid cell has complete spectral and physicochemical attribute data. The time series matrix of spectral reflectance and the soil physicochemical attribute vector are spliced ​​along the feature channel dimension to construct a multi-source heterogeneous data cube containing spectral response features and measured soil salinization indicators. This provides a standardized data input basis for subsequent response consistency analysis, resulting in a patch attribute matrix.

[0025] S102: Based on the attribute matrix of the patch, calculate the ratio of the standard deviation of the spectral reflectance of the patch to the coefficient of variation of the corresponding salinity attribute. Patches with a ratio exceeding a preset threshold are marked as spectral and salinity mismatch units, and the remaining patches are divided into stable response units to generate patch response consistency labels. Based on the patch attribute matrix, for each patch unit corresponding to each row in the matrix, the standard deviation of the spectral reflectance of each band in the entire time series is calculated. The arithmetic mean of the standard deviations of each band is defined as the spectral volatility index. Simultaneously, measured values ​​of total soil salinity and the concentrations of eight major ions are selected, and the ratio of their standard deviations to their arithmetic means is calculated to obtain the coefficient of variation. This coefficient of variation is defined as the salinity attribute dispersion index. A division operation is performed, dividing the spectral volatility index by the salinity attribute dispersion index, to obtain the response ratio characterizing the difference in the rate of change between the two. A response ratio calculated based on statistical distribution characteristics is then established. The critical threshold for consistency is 1.85. Patches with a calculated response ratio exceeding 1.85 are identified as anomalous units whose spectral response is out of sync with salinity changes. These units exhibit drastic fluctuations in salinity while their spectral characteristics are relatively sluggish or opposite. The status attribute field of such patches is marked as "Mismatch". Conversely, patches with a response ratio less than or equal to 1.85 are marked as "Stable". All patch units in the matrix are traversed to complete the status determination and marking operations. A binary label sequence containing spatial index and response status identifier is established to generate patch response consistency labels.

[0026] S103: Based on the consistency label of the patch response, extract the spatial coordinates and attribute features of the mismatched units, aggregate them according to the adjacency relationship to form a continuous region, label the type of mismatch dominant factor, and obtain the salt response anomaly patch set; Based on the consistency label of the map patch response, the spatial geometric coordinates and attribute feature data of all mismatched units marked "Mismatch" are retrieved and extracted. Using a spatial connectivity analysis algorithm, the mismatched units are scanned traversally using eight-neighbor pixels as the search template. Pixels that are spatially adjacent and have the same mismatch label are aggregated into independent connected polygons. The number of pixels in each connected polygon is counted, and its corresponding actual surface area is calculated. A minimum mapping area threshold of 500 square meters is set, and isolated noise patches with areas smaller than this threshold are removed. For each retained continuous region, the mean spectral fluctuation index and the salinity attribute dispersion index of all pixels within the region are calculated. The mean values ​​are compared. If the mean value of the spectral volatility index exceeds 2.0 times the mean value of the salinity property dispersion index, the region is identified as a spectral-dominant mismatch and marked with the identifier "Spectral_Dominant". If the mean value of the salinity property dispersion index exceeds 2.0 times the mean value of the spectral volatility index, it is identified as a salinity-dominant mismatch and marked with the identifier "Salinity_Dominant". If the ratio between the two is between 0.5 and 2.0, it is identified as a mixed mismatch and marked with the identifier "Mixed_Factor". This process completes the identification of the dominant factor type and attribute association for all continuous regions, resulting in a set of salinity response anomaly patches.

[0027] Please see Figure 3 The specific steps for obtaining the spectral correction vector of the patch are as follows: S201: Based on the salt response anomaly patch set, the joint distribution density of normalized vegetation index and soil salinity index of all patch units is statistically analyzed, the boundary contour of the index distribution main cluster is identified, and the patches located outside the contour are uniquely encoded and bound to the patches to form an index offset identifier set. Based on the set of anomalous salt response patches, the spectral reflectance data of each patch unit in the set are traversed. The normalized vegetation index (NDI) is calculated using the reflectance of the near-infrared and red bands, and the soil salinity index is constructed using the reflectance of the blue, red, and near-infrared bands. A two-dimensional feature scatter plane is constructed with the NDI as the x-axis and the soil salinity index as the y-axis. The kernel density estimation algorithm is used to perform probability density mapping on the scatter distribution in the plane. Closed regions of probability density contour lines are identified as the main clusters of the index distribution. The probability density abrupt change points at the edges of the main clusters are extracted and fitted to form closed boundary contours. The projection coordinates of all patches on the two-dimensional plane are scanned one by one. The ray method is used to determine whether the patch coordinates are located outside the boundary contour. Patches with projection coordinates outside the contour are defined as anomalous samples that deviate from the normal response pattern. The unique identification code of this type of patch is extracted and associated with the original spatial index to construct an index list containing the offset patch ID and its coordinate position on the feature plane. The index characteristics and state determination results of the offset patches are recorded in detail to form an index offset identifier set.

[0028] S202: Based on the exponential offset identifier set, determine the azimuth of each offset patch in the exponential plane. If the azimuth points to the high vegetation and low salinity quadrant, it is marked as vegetation interference type. If it points to the low vegetation and high salinity quadrant, it is marked as bare soil enhancement type, thus obtaining the patch interference type classification. Based on the exponential offset identifier set, the geometric centroid coordinates of the main clusters of exponential distribution in the two-dimensional feature plane are calculated. A local polar coordinate system is constructed with this centroid as the origin. Each offset patch in the identifier set is traversed, and the position vector of the patch coordinate point relative to the centroid is calculated. The angle between this position vector and the positive direction of the horizontal axis is determined as the azimuth. The feature plane is divided into four quadrants. The area with an azimuth between 0 and 90 degrees and an NDVI value higher than the centroid and an SI value lower than the centroid is defined as the vegetation disturbance area. The area with an azimuth between 270 and 360 degrees and an NDVI value lower than the centroid and an SI value higher than the centroid is defined as the bare soil enhancement area. Based on the calculated azimuth angle values, the quadrant region to which each patch belongs is determined. If the patch is located in a vegetation interference area, it indicates that the vegetation cover has masked the spectral characteristics of soil salinity, and it is assigned the classification label "Vegetation_Interference". If the patch is located in a bare soil enhancement area, it indicates that the bare soil background has enhanced the salinity signal, and it is assigned the classification label "BareSoil_Enhancement". For patches located in quadrant regions, the label "Undefined_Offset" is assigned. The classification results are written into the attribute table of the patch to obtain the patch interference type classification.

[0029] S203: Based on the classification of patch interference types, the near-infrared band weight is reduced and the short-wave infrared band gain is increased for vegetation-interference patches, and the band weight is adjusted in the opposite direction for bare soil-enhanced patches. The band adjustment amount and direction are recorded and integrated into a patch spectral correction vector. Based on the classification of image patch interference types, for patches labeled "Vegetation_Interference", a vegetation suppression type band weight vector is constructed. The near-infrared band weight coefficient is set to 0.8 to reduce the influence of high vegetation reflectivity, and the shortwave infrared I and shortwave infrared II band weight coefficients are set to 1.2 to enhance water and salt absorption characteristics. For patches labeled "BareSoil_Enhancement", a bare soil correction type band weight vector is constructed. The near-infrared band weight coefficient is set to 1.1, and the shortwave infrared band weight coefficient is set to 0.9 to balance the contribution of soil background brightness to the calculation of salinity index. For undefined types of patches, the full-band weight coefficient is maintained at 1.0. The specially constructed weight coefficient sequence is matched one-to-one with the spectral bands of the patches, and the adjustment amplitude and directional sign of each band are recorded. The adjusted weight configuration parameters are encapsulated into a spectral correction control vector bound to the patch ID to obtain the patch spectral correction vector.

[0030] Please see Figure 4 The specific steps for obtaining the salt feature enhancement gradient distribution dataset are as follows: S301: Based on the spectral correction vector of the patch, perform band weight substitution calculation for each patch, identify the corrected spectral curve, screen patches whose integral performance in the salt-sensitive region of the near-infrared front segment and mid-infrared front segment is lower than the baseline level, and establish a list of feature-weakened patches. Based on the spectral correction vector of the patches, the original full-band spectral reflectance data of each patch is called. The corresponding band weight coefficients are multiplied band by band by band to generate the corrected spectral response curve. The near-infrared front segment with wavelengths of 800 nm to 900 nm and the mid-infrared front segment with wavelengths of 1500 nm to 1600 nm in the spectral curve are locked. The trapezoidal integral method is used to calculate the integral area of ​​the corrected spectral curve in the above two sensitive wavelength segments. The integral area values ​​of the two segments are added to obtain the total salinity sensitivity integral value. The baseline integral threshold obtained from the statistical analysis of the regional background soil spectral characteristics is set to 150.0. Patches with a total salinity sensitivity integral value lower than this threshold are judged as weak feature expression units, indicating that the current weight configuration has not fully highlighted their salinization characteristics. The index information of such patches is extracted to establish a queue to be optimized and a list of feature weakened patches is established.

[0031] S302: Based on the list of feature-weakened patches, for each patch, the weight coefficient of the sensitive band is adjusted step by step with a preset threshold, the spectral reconstruction error and the salt regression residual based on the eight major ion components are calculated, and the error sequence is recorded in the order of adjustment to form error and weight trajectory data. Based on the feature-weakened patch list, patch units within the list are selected, and the weight coefficient of the shortwave infrared I band is locked as the optimization variable. The initial weight value is set to 1.0, the iteration step size is set to 0.05, and the upper limit of the weight search is set to 2.0. Incremental scanning is performed within the weight value range. For each weight step size node, the spectral curve is reconstructed using the current weighted spectral data. The Euclidean distance between the reconstructed curve and the corresponding curve in the standard saline soil spectral library is calculated as the spectral reconstruction error. A prediction model is constructed using the partial least squares regression algorithm with weighted spectral features as independent variables and the measured total amount of eight ions as dependent variables. The sum of squared residuals between the model prediction value and the measured value is calculated as the salinity regression residual. The weight values, spectral reconstruction errors, and salinity regression residuals generated in each iteration are stored sequentially in a three-dimensional array, and the error change trajectory during the optimization process is recorded in detail to form error and weight trajectory data.

[0032] S303: Based on the error and weight trajectory data, calculate the ratio of the decrease in residual to the increase in error between adjacent weight steps. When the ratio exceeds the preset threshold, retain the corresponding weight configuration and combine the patches that meet the conditions and their weight adjustment range to obtain the salt feature enhancement gradient distribution dataset. Based on the error and weight trajectory data, a difference operation is performed on the recorded three-dimensional array sequence to calculate the decrease in salinity regression residuals and the increase in spectral reconstruction error between two adjacent weight steps. The decrease in residuals is divided by the increase in error to obtain the benefit-cost ratio. The benefit-cost ratio threshold is set to 15.0, which represents the minimum accuracy improvement required per unit of spectral distortion. The calculation results are traversed in ascending order of weights. When the benefit-cost ratio of a certain step is lower than 15.0, it is determined that continuing to increase the weights no longer has the cost-effectiveness of optimization. The search is stopped, and the previous weight value of that step is locked as the best enhancement weight for that patch. The patch ID and the best weight value are extracted and combined with the fixed weights of the bands determined in S203 to generate the final patch-specific weight configuration scheme, resulting in the salinity feature enhancement gradient distribution dataset.

[0033] Please see Figure 5 The specific steps for obtaining the weighted structural system for the analysis of salt formation are as follows: S401: Based on the enhanced gradient distribution dataset of saline features, the spectral correction vector of the map patch is spatially registered with the ground-measured molar concentration vectors of eight ions, and the administrative division to which the map patch belongs, the lithology of the parent material, and the irrigation canal system number are superimposed to construct a four-dimensional genetic element record. Based on the enhanced gradient distribution dataset of saline features, this study utilizes the spatial connectivity tool of a geographic information processing platform. Using the coordinates of the geometric center point of the map patch as a reference, the study spatially matches the spectral weight configuration vector specific to the map patch with the molar concentration data of eight ions from ground-based measurement stations. For map patches without direct measurement points, the nearest neighbor method is used to associate them with the nearest measured data. An administrative division layer in vector format is overlaid to extract the administrative jurisdiction unit code to which the map patch belongs. A geological parent material layer is overlaid to extract the sediment type or lithological properties of the location of the map patch. An irrigation canal system vector network layer is overlaid to calculate the Euclidean distance from the center of the map patch to the nearest first-level water conveyance channel and extract the design flow rate and seepage prevention level attributes of the channel. The spectral weight parameters, ion concentration data, administrative division information, parent material category, and water conservancy facility parameters are integrated into a unified data record structure to construct a four-dimensional genetic element record.

[0034] S402: Based on the four-dimensional genetic element record, the co-occurrence frequency of bicarbonate and sodium ions, the antagonistic ratio of sulfate and calcium and magnesium ions, and the correlation slope between chloride ions and conductivity are statistically analyzed. The dominant combination type is divided according to the strength of ion interaction, and the dominant combination type and intervention level are obtained. The co-occurrence frequency of bicarbonate and sodium ions refers to the frequency at which bicarbonate and sodium ions appear simultaneously within the target spatial range by comparing the spectral data of the map patch with the ground-based measured data. The antagonistic ratio between sulfate and calcium and magnesium ions refers to the ratio of their antagonistic interactions in a patch by analyzing the ion concentration data of the patch. The slope of the correlation between chloride ion concentration and conductivity refers to the slope between chloride ion concentration and conductivity obtained through regression analysis of the relationship between chloride ion concentration and conductivity. Based on the four-dimensional causal element records, stoichiometric characteristic analysis was performed on the ion concentration data in each record. The proportion of samples with bicarbonate concentration greater than 2.0 mmol / L and sodium ion concentration greater than 10.0 mmol / L was used as the co-occurrence frequency of bicarbonate and sodium ions. The ratio of sulfate ion concentration to the sum of calcium and magnesium ion concentrations was calculated as the antagonistic ratio between sulfate and calcium and magnesium ions. A linear regression model of chloride ion concentration and soil conductivity was established using samples from the same irrigation area, and the slope was extracted as the correlation slope between chloride ions and conductivity. Based on the above statistical indicators, the salinization causal type was determined. If the co-occurrence frequency was higher than 60% and the antagonistic ratio was less than 1.0, it was determined to be the soda-alkalization dominant type. If the correlation slope was higher than 0.8 and the antagonistic ratio was greater than 1.0, it was determined to be the chloride-strong salinization dominant type. If the antagonistic ratio was between 0.8 and 1.2, it was determined to be the sulfate-neutral salinization dominant type. At the same time, the intervention level was divided into three levels: mild, moderate, and severe, based on the total ion concentration, to obtain the dominant combination type and intervention level.

[0035] S403: Based on the dominant combination type and intervention level, extract the basic weights of eight ion distributions, set the soil parent material distribution adjustment weights and irrigation water source distribution correction weights, link and integrate the three types of weights and determine the level of action to obtain the weighted structure system for salinity cause analysis. Based on the dominant combination type and intervention level, the salinization causation expert knowledge base was invoked to assign basic weight coefficients to the eight ions. For the soda-alkalization dominant type, carbonate and bicarbonate were assigned higher weights, and for the chloride dominant type, chloride and sodium ions were assigned higher weights. Adjustment weights were introduced based on the type of parent material. For sandy parent material with high permeability, the weight allocation of soluble ions was increased. Correction weights were introduced based on the distance to the irrigation water source. For patches adjacent to non-seepage-proof channels, the weight allocation of sulfate ions was increased. The analytic hierarchy process (AHP) was used to perform a weighted summation of the basic weights, adjustment weights, and correction weights, and the resulting vector was normalized to ensure that the sum of all ion weights was 1.0. Based on the final weight distribution, the decisive causal factors were identified and their levels of influence were determined, resulting in a weighted structure system for salinization causation analysis.

[0036] Please see Figure 6 The specific steps for obtaining the salting risk zone layer are as follows: S501: Based on the weighted structure system of salinization cause analysis, the total salt equivalent of eight ions is calculated for each patch, and a salinization index is generated by combining the pH value correction term. The salinization index is compared with the preset threshold boundaries of mild, moderate and severe salinity to obtain the interval assignment result. Based on the weighted structural system of salt formation analysis, each patch unit is extracted. The corresponding measured concentrations of the eight major ions (in Indicates the first The ions (in millimoles per liter) and the normalized weights determined in S403 At the same time, the measured values ​​of soil pH for the plot were obtained. The comprehensive salinity index of the map patch is calculated using the following formula. : In this calculation process, taking a typical patch P_008 as an example, the ion concentration data of the patch is first retrieved, assuming the sodium ion concentration is... The chloride ion concentration was 12.0 mmol / L. The sulfate concentration was 10.0 mmol / L. The concentration of sodium ion is 6.0 mmol / L, and the sum of the concentrations of the other five ions is 5.0 mmol / L; the corresponding normalized weights are retrieved, assuming the sodium ion weight... The chloride ion weight is 0.25. The sulfate weight is 0.30. The weighted average contribution coefficient of the remaining ions is 0.20, and the weighted sum of the weighted ions is converted to 0.25. The first part of the summation operation is then performed: This value is the weighted total salt equivalent; subsequently, the measured soil pH data for this patch were retrieved, assuming... Perform the second part of the correction calculation: Finally, the two results are multiplied together to obtain the comprehensive salinity index. The calculated results The numerical values ​​were compared with preset grading standards, setting the threshold for mild salinization at 2.5, moderate salinization at 5.0, and severe salinization at 8.0. Therefore, it was determined that the patch belongs to the heavily salinized region, and the region assignment result was obtained.

[0037] S502: Based on the interval attribution results, assign a corresponding risk category code to each map patch, merge similar map patches according to spatial adjacency, and obtain a salinization risk partitioning layer; Based on the interval assignment results, the determined risk levels are converted into corresponding integer codes. Non-salinization is set to 0, mild to 1, moderate to 2, and severe to 3. A risk-coded raster matrix is ​​generated by traversing the entire monitoring area. Raster reclassification and aggregation algorithms are applied, using four-neighbor or eight-neighbor connections as the connectivity condition, to merge spatially adjacent raster units with the same code into a single polygon object. A mode filtering operation is performed to eliminate isolated single-pixel noise points. For small patches with an area less than 100 square meters, their attributes are merged into the adjacent patch with the longest common boundary. This completes the fragmentation and smoothing process. A unique spatial ID is assigned to each merged polygon area, and the corresponding risk level attribute data is attached. A vector-format geospatial data layer is constructed to obtain the salinization risk zoning layer.

[0038] Please see Figure 7 A GIS-based spatial analysis system for salinization includes: The spectral and salinity mismatch identification module is based on remote sensing image patch units. It extracts the spectral reflectance sequence of the patches within the current monitoring period, obtains the total soil salinity, eight major ion components and pH value of the corresponding patch location, and identifies the mismatch area between the spectral and salinity relationship by comparing the fluctuation amplitude of the patch spectral response with the dispersion of salinity attributes, thus obtaining a set of abnormal salinity response patches. The spectral correction factor generation module is based on the salt response anomaly patch set. It analyzes the joint distribution of normalized vegetation index and soil salinity index of patches in high-throughput remote sensing patch units, compares the offset direction of the main distribution cluster boundary corresponding to the index points, reduces the weight of vegetation sensitive band for patches with upper boundary offset, and records the adjustment content for patches with lower boundary offset to obtain the patch spectral correction vector. The salt feature gradient calculation module performs band weight substitution calculation on the patches based on the patch spectral correction vector, filters patches whose salt-sensitive band response is lower than the threshold after correction, increases the weight by a fixed step size and records the spectral reconstruction error and salt fitting residual, calculates the error descent gradient, and obtains the salt feature enhancement gradient distribution dataset. The causal weight fusion module is based on the salinity feature enhanced gradient distribution dataset. It spatially registers the spectral correction vector of the map patch with the measured data of eight major ions, and superimposes geographical location, soil parent material type and irrigation water source type. Based on the ion co-occurrence frequency and the geographical intervention level, it obtains the weighted structure system for salinity causal analysis. The risk zoning generation module is based on the weighted structure system of salinity cause analysis. It calculates the salinity index of the map patches, classifies the index with the threshold ranges of mild, moderate and severe, assigns risk category codes according to the judgment results, merges adjacent map patches, and outputs the salinized risk zoning layer.

[0039] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A GIS-based spatial analysis method for salinization, characterized in that, Includes the following steps: S1: Based on remote sensing image patch units, extract the spectral reflectance sequence of patches within the current monitoring period, identify the measured total soil salinity, content of eight major ion components and pH value of the corresponding patch location, compare spectral fluctuations with salt separation divergence, identify mismatch areas, and obtain a set of abnormal salinity response patches. S2: Based on the set of anomalous patches in salinity response, analyze the high-throughput patch index, compare the offset direction of the normalized vegetation index and the soil salinity index, set correction factors for patches that deviate from the upper and lower boundaries of the main distribution cluster, and obtain the patch spectral correction vector. S3: Based on the spectral correction vector of the patch, the original spectrum is calculated, and the patch is determined to be in a weakened state when the corrected spectrum is lower than the preset threshold. The salt sensitivity weight is increased by step size, and the reconstruction error and salt fitting residual after adjustment are recorded to obtain the salt feature enhancement gradient distribution dataset. S4: Based on the enhanced gradient distribution dataset of salinity features, combined with ground-measured ion data, the geographical location, parent material, irrigation water source and salt ion combination are weighted and fused. According to the contribution of factors to the salinity formation mechanism, the level of influence is assigned to obtain the weighted structure system for salinity causation analysis.

2. The GIS-based spatial analysis method for salinization according to claim 1, characterized in that, The set of anomalous patches in the salinity response includes anomalous patch number, spatial clustering range, and mismatch type. The patch spectral correction vector includes correction factor value, adjustment band identifier, and patch clustering label. The salinity feature enhancement gradient distribution dataset includes gradient direction, gradient amplitude range, and response enhancement level. The weighted structure system for salinity causation analysis includes ion combination weight, geographical factor level, and water source influence order.

3. The GIS-based spatial analysis method for salinization according to claim 1, characterized in that, The specific steps for obtaining the salt-response anomaly patch set are as follows: S101: Based on remote sensing image patch units, extract the time series of reflectance of patches in the visible to shortwave infrared bands within the current monitoring period, and simultaneously match the measured records of total soil salinity, eight major ion components and pH value at the same patch location to identify the multidimensional attribute vector of each patch and obtain the patch attribute matrix. S102: Based on the attribute matrix of the patch, calculate the ratio of the standard deviation of the spectral reflectance of the patch to the coefficient of variation of the corresponding salinity attribute. Patches with a ratio exceeding a preset threshold are marked as spectral and salinity mismatch units, and the remaining patches are divided into stable response units to generate patch response consistency labels. S103: Based on the consistency label of the patch response, extract the spatial coordinates and attribute features of the mismatched units, aggregate them according to the adjacency relationship to form a continuous region, label the mismatch dominant factor type, and obtain the salt response abnormal patch set.

4. The GIS-based spatial analysis method for salinization according to claim 3, characterized in that, The specific steps for obtaining the spectral correction vector of the patch are as follows: S201: Based on the set of anomalous patches in response to salinity, the joint distribution density of normalized vegetation index and soil salinity index of all patch units is statistically analyzed, the boundary contour of the main cluster of index distribution is identified, and the patches located outside the contour are uniquely encoded and bound to the patches to form an index offset identifier set. S202: Based on the index offset identifier set, determine the azimuth angle of each offset patch in the index plane. If the azimuth angle points to the high vegetation and low salinity quadrant, it is marked as vegetation interference type. If it points to the low vegetation and high salinity quadrant, it is marked as bare soil enhancement type, thus obtaining the patch interference type classification. S203: Based on the classification of the interference types of the patches, reduce the weight of the near-infrared band and increase the gain of the short-wave infrared band for vegetation interference patches, and reverse the band weight for bare soil enhancement patches. Record the band adjustment amount and direction, and integrate them into a patch spectral correction vector.

5. The GIS-based spatial analysis method for salinization according to claim 4, characterized in that, The specific steps for obtaining the salt feature enhancement gradient distribution dataset are as follows: S301: Based on the spectral correction vector of the patch, perform band weight substitution calculation on each patch, identify the corrected spectral curve, screen patches whose integral performance in the salt-sensitive region of the near-infrared front segment and the mid-infrared front segment is lower than the benchmark level, and establish a list of feature-weakened patches. S302: Based on the list of weakened feature patches, for each patch, the weight coefficient of the sensitive band is adjusted step by step with a preset threshold, the spectral reconstruction error and the salt regression residual based on the eight major ion components are calculated, and the error sequence is recorded in the order of adjustment to form error and weight trajectory data. S303: Based on the error and weight trajectory data, calculate the ratio of the decrease in residual to the increase in error between adjacent weight steps. When the ratio exceeds a preset threshold, retain the corresponding weight configuration and combine the map features that meet the conditions and their weight adjustment ranges to obtain the salt feature enhancement gradient distribution dataset.

6. The GIS-based spatial analysis method for salinization according to claim 5, characterized in that, The specific steps for obtaining the weighted structural system for the analysis of salt formation are as follows: S401: Based on the salinity feature enhancement gradient distribution dataset, spatially register the spectral correction vector of the patch with the ground-measured molar concentration vectors of the eight ions, and superimpose the administrative division to which the patch belongs, the parent material lithology category, and the irrigation canal system number to construct a four-dimensional genetic element record. S402: Based on the four-dimensional causal element record, the co-occurrence frequency of bicarbonate and sodium ions, the antagonistic ratio of sulfate and calcium and magnesium ions, and the correlation slope between chloride ions and conductivity are statistically analyzed. The dominant combination type is divided according to the intensity of ion interaction, and the dominant combination type and intervention level are obtained. S403: Based on the dominant combination type and intervention level, extract the basic weights of eight ion distributions, set the soil parent material distribution adjustment weights and irrigation water source distribution correction weights, link and integrate the three types of weights and determine the level of action to obtain the weighted structure system for salinization cause analysis.

7. The GIS-based spatial analysis method for salinization according to claim 6, characterized in that, The co-occurrence frequency of bicarbonate and sodium ions refers to the frequency at which bicarbonate and sodium ions appear simultaneously within the target spatial range by comparing the spectral data of the map patch with the ground-based measured data. The antagonistic ratio between sulfate and calcium and magnesium ions refers to the ratio of their antagonism in the patch obtained by analyzing the ion concentration data of the patch and analyzing the interaction strength between sulfate and calcium and magnesium ions. The slope of the correlation between chloride ions and conductivity refers to the slope between chloride ion concentration and conductivity obtained by regression analysis of the relationship between chloride ion concentration and conductivity.

8. The GIS-based spatial analysis method for salinization according to claim 1, characterized in that, The method also includes step S5: S5: Based on the weighted structure system of the salinity analysis, the salinity evaluation value of the map patch is multiplied by the level of action. The product result is then assigned to the threshold range of mild, moderate and severe salinization. The map patch risk categories are divided according to the determination results to obtain the salinization risk zoning layer. The salinization risk zoning layer includes risk category code, threshold attribution range, and patch spatial index.

9. The GIS-based spatial analysis method for salinization according to claim 8, characterized in that, The specific steps for obtaining the salinization risk zoning layer are as follows: S501: Based on the weighted structure system for the analysis of salinity formation, calculate the total salt equivalent of eight ions for each patch, generate a salinity index by combining the pH correction term, and compare the salinity index with the preset thresholds of mild, moderate and severe salinity to obtain the interval assignment result. S502: Based on the interval attribution results, assign a corresponding risk category code to each map patch, merge similar map patches according to spatial adjacency, and obtain a salinization risk zoning layer.

10. A GIS-based spatial analysis system for salinization, characterized in that, The system is used to implement the GIS-based spatial analysis method for salinization as described in any one of claims 1-9, and the system comprises: The spectral and salinity mismatch identification module is based on remote sensing image patch units. It extracts the spectral reflectance sequence of the patches within the current monitoring period, obtains the total soil salinity, eight major ion components and pH value of the corresponding patch location, and identifies the mismatch area between the spectral and salinity relationship by comparing the fluctuation amplitude of the patch spectral response with the dispersion of salinity attributes, thus obtaining a set of abnormal salinity response patches. The spectral correction factor generation module analyzes the joint distribution of normalized vegetation index and soil salinity index of the patches in the high-throughput remote sensing patch unit based on the set of salinity response anomalies. It compares the offset direction of the boundary of the main distribution cluster corresponding to the index points, reduces the weight of vegetation sensitive bands for patches with upper boundary offset, and records the adjustment content for patches with lower boundary offset, thus obtaining the spectral correction vector of the patch. Based on the spectral correction vector of the patch, the salt feature gradient calculation module performs band weight substitution operation on the patch, filters the patches whose salt-sensitive band response is lower than the threshold after correction, increases the weight by a fixed step size and records the spectral reconstruction error and salt fitting residual, calculates the error descent gradient, and obtains the salt feature enhancement gradient distribution dataset. The causal weight fusion module, based on the salinity feature enhancement gradient distribution dataset, spatially registers the patch spectral correction vector with the measured data of eight ions, overlays geographical location, soil parent material type and irrigation water source type, and obtains a weighted structure system for salinity causal analysis based on ion co-occurrence frequency and geographical intervention level. The risk zoning generation module calculates the salinity index of the map patches based on the weighted structure system of the salinity cause analysis, classifies the index with the threshold ranges of mild, moderate and severe, assigns risk category codes according to the judgment results, merges adjacent map patches, and outputs the salinization risk zoning layer.

Images