High and steep slope ecological restoration evaluation method and system based on multi-source data fusion

Abstract

The application relates to the technical field of data analysis, and provides a high and steep slope ecological restoration evaluation method and system based on multi-source data fusion. Embodiments of the application utilize multi-source observation data of a high and steep slope region to form a slope restoration observation data set with time sequence correlation, which contains surface coverage spectrum, soil environment monitoring, slope surface image and vegetation physiological characteristic data. According to the data set, a correlation mapping relationship between restoration indexes and environmental factors is generated, and a multi-dimensional evaluation model is constructed. A time sequence analysis algorithm is used to dynamically track the trend of key indexes, and a fuzzy comprehensive evaluation method is combined to obtain a comprehensive evaluation result. Based on the result, a visual evaluation report containing a time axis evolution trend graph, a spatial distribution heat map and a restoration grade division boundary line is generated, and is pushed to an ecological restoration management platform to support restoration strategy adjustment decision-making.

Description

Technical Field

[0001] This application belongs to the field of data analysis technology, specifically relating to a method and system for evaluating the ecological restoration of steep slopes based on multi-source data fusion. Background Technology

[0002] Ecological restoration of steep slopes is of great significance for maintaining ecological balance, ensuring infrastructure safety, and improving the quality of the regional ecological environment. Early research in this field focused primarily on engineering measures, such as retaining walls and slope protection, to enhance slope stability. However, these measures often overlooked the self-repair and sustainable development of the ecosystem. With the development of ecological concepts, ecological restoration methods such as vegetation restoration have gradually been introduced.

[0003] However, the aforementioned existing technologies suffer from limitations such as a single data source and incomplete evaluation dimensions, failing to comprehensively and accurately reflect the actual situation of ecological restoration of steep slopes. Furthermore, existing technologies lack effective tracking and evaluation of the dynamic changes in the ecological restoration process, resulting in insufficient timeliness and practicality of the evaluation results, making it difficult to provide a scientific and comprehensive basis for adjusting ecological restoration strategies. Summary of the Invention

[0004] This application provides a method and system for evaluating the ecological restoration of steep slopes based on multi-source data fusion.

[0005] In a first aspect, embodiments of this application provide a method for evaluating the ecological restoration of steep slopes based on multi-source data fusion, applied to a system for evaluating the ecological restoration of steep slopes. The method includes: Using multi-source observation data from high and steep slope areas, a slope restoration observation dataset with time-series correlation is formed; wherein, the slope restoration observation dataset includes land cover spectral data collected by remote sensing satellites, soil environmental monitoring data collected by deploying sensor networks, slope surface image data collected by UAV platforms, and vegetation physiological characteristic data obtained through field sampling. Based on the slope restoration observation dataset, a correlation mapping relationship between restoration indicators and environmental factors is generated. Based on the correlation mapping relationship, a multi-dimensional evaluation model including vegetation growth status, soil physicochemical properties, and slope structural stability is constructed. The time series analysis algorithm is used to track the dynamic trends of vegetation coverage, soil shear strength and slope erosion rate in the multi-dimensional evaluation model over a continuous period of time, generating index change tracking information. The fuzzy comprehensive evaluation method is then used to quantify the restoration effect level of the index change tracking information, and a comprehensive evaluation result of the ecological restoration effect of high and steep slopes is obtained. Based on the comprehensive evaluation results, a visual assessment report is generated, which includes a time-axis evolution trend map, a spatial distribution heat map, and boundary lines for the classification of restoration levels. The visual assessment report is then pushed to the ecological restoration management platform to support the decision-making on the adjustment of restoration strategies.

[0006] Secondly, this application provides an ecological restoration evaluation system for steep slopes, which includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the above-described method.

[0007] Thirdly, embodiments of this application provide a computer-readable storage medium including a computer program, which, when run on a high and steep slope ecological restoration evaluation system, causes the high and steep slope ecological restoration evaluation system to perform the steps of the above-described method. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating an evaluation method for ecological restoration of steep slopes based on multi-source data fusion, provided in an embodiment of this application.

[0009] Figure 2 This is a schematic diagram of the structure of an ecological restoration evaluation system for steep slopes provided in an embodiment of this application. Detailed Implementation

[0010] See Figure 1 This is a method for evaluating the ecological restoration of steep slopes based on multi-source data fusion provided in this application embodiment. This method can be applied to the ecological restoration evaluation system of steep slopes. The specific process is as follows: steps 110-140.

[0011] Step 110: Utilize multi-source observation data from steep slope areas to form a slope restoration observation dataset with time-series correlation; wherein, the slope restoration observation dataset includes land cover spectral data collected by remote sensing satellites, soil environmental monitoring data collected by deploying a sensor network, slope surface image data collected by a drone platform, and vegetation physiological characteristic data obtained through on-site sampling.

[0012] In this embodiment of the application, when remote sensing satellites collect surface cover spectral data, they scan steep slope areas at different times. Due to the influence of factors such as satellite orbit, lighting conditions, and atmospheric conditions, the data collected each time has certain characteristics. For example, data collected by satellites may be clearer and more accurate during a clear day, but data quality may be affected under cloudy or hazy weather conditions.

[0013] The deployment of the sensor network involves installing various sensors at different locations in the steep slope area. These sensors continuously monitor various environmental parameters of the soil. For example, the soil moisture, pH, and nutrient content may differ at different locations such as the top, middle, and bottom of the slope, and the sensors record these data in real time.

[0014] The drone platform will then photograph the slope surface according to a preset flight route and altitude, acquiring high-resolution image data of the slope surface. On-site sampling is carried out by professionals who regularly collect vegetation samples from steep slope areas. The physiological characteristics of the vegetation, such as leaf size and chlorophyll content, are obtained through laboratory analysis.

[0015] Each of the aforementioned multi-source observation data has unique value. By associating them according to their collection time, a complete time-series slope restoration observation dataset can be formed.

[0016] In one implementation, step 110 includes: Step 111: Based on the interference of changes in illumination conditions on spectral reflectance, perform atmospheric correction and radiation normalization on the surface cover spectral data to obtain the target spectral data.

[0017] In areas with steep slopes, lighting conditions vary with time, weather, and geographical location, which significantly interferes with the spectral reflectance of surface cover spectral data. Atmospheric components such as water vapor and aerosols absorb and scatter light, preventing satellite-collected spectral data from accurately reflecting the true surface conditions.

[0018] To eliminate this interference, atmospheric correction is required. The atmospheric correction process begins by establishing an atmospheric model based on the meteorological conditions and atmospheric parameters at the time of satellite data acquisition. This model simulates the influence of the atmosphere on light, and then the acquired spectral data is inversely corrected to remove atmospheric interference. For example, in hazy weather, aerosols in the atmosphere reduce the overall reflectance of the spectral data; atmospheric correction can restore this reduced reflectance to a level close to the actual reflectance of the Earth's surface.

[0019] Radiometric normalization is used to make spectral data collected at different times and under different conditions comparable. Due to factors such as satellite orbit and sensor sensitivity, spectral data collected at different times may have radiometric differences. Radiometric normalization selects a standard radiometric reference and adjusts all collected spectral data according to this reference to unify their radiometric levels. The target spectral data obtained after this processing can more accurately reflect the true spectral characteristics of the land cover in steep slope areas.

[0020] Step 112: By extrapolating the trend of monitoring values ​​at adjacent time points, outlier identification and missing value interpolation are performed on the soil environmental monitoring data to obtain continuous time-series soil environmental data.

[0021] Optionally, soil environmental monitoring data is collected in real time through a sensor network. However, during the collection process, factors such as sensor malfunctions and communication interference may affect the data, leading to outliers or missing values. To obtain continuous and accurate soil environmental data, these outliers and missing values ​​need to be processed. The trend extrapolation method for monitoring values ​​at adjacent time points is based on the principle that soil environmental parameters have a certain continuity and trend of change over a short period of time. For outlier identification, the relationship between each monitoring value and the monitoring values ​​at adjacent time points is analyzed. If a monitoring value differs too much from the monitoring value at an adjacent time point, exceeding the normal range of change, it is determined to be an outlier. For example, under normal circumstances, the change in soil moisture at adjacent time points should be relatively stable. If the soil moisture at a certain time point suddenly jumps significantly, it may be an outlier. For the imputation of missing values, the missing value is predicted based on the monitoring values ​​at adjacent time points and their changing trends. If soil moisture data is missing at a certain time point, the soil moisture monitoring values ​​before and after that time point are observed, their changing trends are analyzed, and then the missing soil moisture value is estimated based on this trend. This method can correct outliers and fill in missing values ​​to obtain continuous time-series soil environmental data.

[0022] Step 113: Based on the coordinates of ground control points, perform distortion correction and spatial registration processing on the slope surface image data of different flight periods to obtain spatially aligned surface image data.

[0023] Slope surface images acquired by drones at different flight times may suffer from distortion and spatial inconsistencies due to factors such as flight attitude and camera distortion. To address these issues, distortion correction and spatial registration based on ground control point coordinates are necessary. Ground control points are pre-defined points with precise geographical coordinates in steep slope areas. The positions of these ground control points in the image are recorded simultaneously during slope surface image acquisition. Distortion correction first analyzes the camera's distortion model and then performs an inverse transformation on the image based on this model to eliminate distortion caused by the camera itself. For example, camera lenses may cause barrel or pincushion distortion; distortion correction can restore these distortions to normal. Spatial registration involves spatially aligning images acquired at different flight times. Using the coordinates of the ground control points as a reference, the spatial transformation relationship between different images is calculated. Then, the images are translated, rotated, and scaled to align them spatially. The resulting spatially aligned surface image data accurately reflects the surface condition of steep slope areas.

[0024] Step 114: Perform sample verification on the vegetation physiological characteristic data to obtain the target physiological characteristic data.

[0025] Errors or inaccuracies may occur during the collection and recording of vegetation physiological characteristic data obtained from field sampling. To ensure the accuracy and reliability of the data, sample verification is necessary. This process includes reconfirming the information of the collected vegetation samples, checking the accuracy of the collection location, collection time, sample number, etc. Simultaneously, the vegetation physiological characteristic data obtained from laboratory analysis will be reviewed to check the correctness of the measurement methods and the reasonableness of the results. For example, if the measured leaf size data deviates significantly from the normal leaf size range for that plant, remeasurement or a review of the measurement method is required. Through sample verification, erroneous or inaccurate data can be eliminated, resulting in the obtained target physiological characteristic data.

[0026] Step 115: Perform time series correlation processing on the target spectral data, the continuous temporal soil environmental data, the spatially aligned surface image data, and the target physiological characteristic data according to the collection timestamp to generate a slope restoration observation dataset containing data collection timestamps.

[0027] After preprocessing various data types, the target spectral data, continuous time-series soil environmental data, spatially aligned surface image data, and target physiological characteristic data need to be integrated to form a slope restoration observation dataset with time-series correlation. Time-series correlation processing based on collection timestamps associates each data point with its collection time. For example, for a spectral value in the target spectral data, the specific time it was collected is recorded; for a soil moisture value in the continuous time-series soil environmental data, its collection time is also recorded. In this way, all data are arranged in chronological order to form a complete dataset. Simultaneously, a data collection time stamp is added to each data point, allowing for convenient data filtering and comparison based on time in subsequent analysis. The final generated slope restoration observation dataset contains multi-source observation data from steep slope areas at different time points.

[0028] Step 120: Generate a correlation mapping relationship between the restoration index and environmental factors based on the slope restoration observation dataset, and construct a multi-dimensional evaluation model based on the correlation mapping relationship, including the vegetation growth status dimension, soil physicochemical property dimension, and slope structural stability dimension.

[0029] In this embodiment, to accurately evaluate the ecological restoration effect of steep slopes, multiple dimensions need to be considered. Through in-depth analysis of the slope restoration observation dataset, a correlation mapping relationship between restoration indicators and environmental factors is generated, thereby constructing a multi-dimensional evaluation model. Restoration indicators reflect the specific situation of ecological restoration of steep slopes, such as vegetation coverage and soil shear strength; environmental factors are various factors affecting restoration indicators, such as sunlight and soil moisture. The correlation mapping relationship describes the interrelationship between restoration indicators and environmental factors. By analyzing this relationship, it is possible to understand which environmental factors have a greater impact on restoration indicators. Based on this correlation mapping relationship, a multi-dimensional evaluation model is constructed, including the vegetation growth status dimension, soil physicochemical properties dimension, and slope structural stability dimension. The vegetation growth status dimension focuses on the growth of vegetation, such as vegetation coverage and biomass; the soil physicochemical properties dimension examines the physical and chemical properties of the soil, such as soil moisture and pH; and the slope structural stability dimension assesses the structural stability of the slope, such as the slope erosion rate. Multidimensional evaluation models can comprehensively consider multiple factors and more accurately evaluate the ecological restoration effect of steep slopes.

[0030] As an optional embodiment, step 120 includes: Step 121: Extract the first observation variable related to vegetation growth status from the slope restoration observation dataset, and use the first observation variable as the first candidate restoration index for the vegetation growth status dimension.

[0031] The slope restoration observation dataset contains a wealth of information related to vegetation growth status. Through screening and analysis of this data, the primary observation variable was extracted, which may include vegetation height, density, and chlorophyll content. Vegetation height reflects the overall growth level of the vegetation; higher vegetation height generally indicates better growth. Vegetation density reflects the amount of vegetation per unit area; higher density may indicate better vegetation growth. Chlorophyll content is related to the photosynthetic capacity of the vegetation; higher chlorophyll content helps the vegetation to perform more effective photosynthesis. These primary observation variables were chosen as the first candidate restoration indicators for the vegetation growth status dimension because they can reflect the vegetation's growth status from different perspectives.

[0032] Step 122: Extract a second observation variable related to soil physicochemical properties from the slope restoration observation dataset, and use the second observation variable as the second candidate restoration index for the soil physicochemical properties dimension.

[0033] A second observational variable related to soil physicochemical properties was extracted from the slope restoration observation dataset. This variable may include soil moisture, pH, and nutrient content. Soil moisture affects vegetation growth and soil physical properties; suitable soil moisture promotes root growth and water absorption. Soil pH affects nutrient availability and microbial activity; different vegetation types have different tolerance ranges for soil pH. Soil nutrient content is the material basis for vegetation growth; sufficient nutrients promote vegetation growth and development. Using these second observational variables as the second candidate restoration index for soil physicochemical properties can accurately reflect the state of soil physicochemical properties.

[0034] Step 123: Extract the third observation variable related to slope structural stability from the slope restoration observation dataset, and use the third observation variable as the third candidate restoration index for the slope structural stability dimension.

[0035] A third observation variable related to slope structural stability was extracted from the slope restoration observation dataset. This variable may include slope erosion rate and soil shear strength. Slope erosion rate reflects the speed at which the soil on the slope surface is eroded. An excessively rapid erosion rate leads to soil loss from the slope surface, affecting vegetation growth and slope stability. Soil shear strength reflects the soil's ability to resist shear forces; higher soil shear strength helps maintain slope structural stability. Using this third observation variable as a third candidate restoration indicator for the slope structural stability dimension can effectively assess the slope's structural stability and provide a basis for determining whether there are potential safety hazards.

[0036] Step 124: Based on the statistical correlation between each candidate remediation index and its corresponding environmental factor, generate the correlation coefficient matrix between the index and the factor through bivariate correlation analysis.

[0037] Bivariate correlation analysis is a method used to study the linear relationship between two variables. In this embodiment, bivariate correlation analysis is performed on each candidate restoration indicator and its corresponding environmental factor to calculate the correlation coefficient between them. For example, for a candidate restoration indicator in the vegetation growth status dimension, such as vegetation cover, the relationship between it and light intensity in the environmental factor is analyzed. By collecting a large amount of vegetation cover and light intensity data, bivariate correlation analysis is used to calculate the correlation coefficient between them. This coefficient represents the degree of linear correlation between vegetation cover and light intensity. By arranging the correlation coefficients of all candidate restoration indicators and environmental factors in a certain way, a correlation coefficient matrix between indicators and factors can be generated. This matrix can intuitively show the correlation between each candidate restoration indicator and environmental factor.

[0038] Preferably, step 124 includes: Step 1241: Synchronize and match the time series data of each candidate repair indicator with the corresponding environmental factor time series data to obtain a time series synchronization matching dataset.

[0039] Before conducting bivariate correlation analysis, it is necessary to ensure that the time-series data of candidate restoration indicators and their corresponding environmental factors are temporally correlated. Synchronizing and matching the time-series data of each candidate restoration indicator with its corresponding environmental factor aligns them according to their collection time. For example, for a candidate restoration indicator in the vegetation growth status dimension, such as the time-series data of vegetation height, and the environmental factor data of temperature, their data at the same time points will be matched. If vegetation height data is available at a certain time point but no corresponding temperature data, or vice versa, data filtering or interpolation is required to ensure that each time point has corresponding candidate restoration indicator data and environmental factor data. The final time-series synchronized matching dataset contains the data of each candidate restoration indicator and its corresponding environmental factor at the same time points.

[0040] Step 1242: Perform stationarity verification on the time-series synchronization matching dataset. If the data does not meet the stationarity requirements, perform differencing until the time-series synchronization matching dataset passes the stationarity verification.

[0041] Stationarity is a crucial concept in time series analysis. When performing bivariate correlation analysis, the time-series synchronous matching dataset is required to be stationary. Stationarity means that the statistical properties of the time series do not change over time; for example, the mean and variance remain stable. Validating the stationarity of the time-series synchronous matching dataset involves using statistical tests to determine whether the data meets the stationarity requirement. If the data does not meet the stationarity requirement, differencing is necessary. Difference involves calculating the difference between adjacent time points in the time series data. For example, for a time series dataset, subtracting the data from the previous time point from the data at the next time point yields a new time series. This differencing process is repeated until the dataset passes the stationarity validation. Data that has undergone stationarity validation and differencing better meets the requirements of bivariate correlation analysis, improving the accuracy of the analysis results.

[0042] Step 1243: Perform bivariate correlation analysis on the time-series synchronous matching dataset that has passed the stationarity verification to obtain the Pearson correlation coefficient; wherein, the Pearson correlation coefficient is used to represent the degree of linear association between candidate remediation indicators and environmental factors.

[0043] After the time-series synchronous matching dataset passes stationarity verification, bivariate correlation analysis can be performed. The Pearson correlation coefficient is a commonly used indicator to measure the degree of linear association between two variables. For each candidate remediation indicator and its corresponding environmental factor in the stationarity-verified time-series synchronous matching dataset, the Pearson correlation coefficient is calculated using bivariate correlation analysis. This coefficient ranges from -1 to 1. If the Pearson correlation coefficient is close to 1, it indicates a strong positive linear association between the candidate remediation indicator and the environmental factor; that is, when one variable increases, the other also increases. If it is close to -1, it indicates a strong negative linear association; that is, when one variable increases, the other decreases. If it is close to 0, it indicates a weak linear association. By calculating the Pearson correlation coefficient, the degree of linear association between the candidate remediation indicator and the environmental factor can be quantified.

[0044] Step 1244: Arrange the Pearson correlation coefficients according to the cross-combination of candidate remediation indicators and environmental factors to generate a preliminary correlation coefficient matrix.

[0045] The calculated Pearson correlation coefficients between each candidate restoration index and environmental factor are arranged according to the cross-combination of the candidate restoration index and environmental factor. For example, for multiple candidate restoration indices and multiple environmental factors in the vegetation growth status dimension, the Pearson correlation coefficients of each candidate restoration index and each environmental factor are arranged into a matrix. The rows of the matrix represent candidate restoration indices, the columns represent environmental factors, and each element in the matrix is ​​the corresponding Pearson correlation coefficient between the candidate restoration index and the environmental factor. The preliminary correlation coefficient matrix generated in this way can intuitively show the degree of linear correlation between each candidate restoration index and environmental factor.

[0046] Step 1245: Perform a significance level check on the preliminary correlation coefficient matrix, remove correlation coefficients that fail the significance check, retain correlation coefficients that pass the significance check, and generate a correlation coefficient matrix of indicators and factors.

[0047] The Pearson correlation coefficients in the initial correlation coefficient matrix may contain some randomness, requiring significance level verification to ensure the reliability of the results. Significance level verification uses statistical tests to determine whether each Pearson correlation coefficient is statistically significant. If a Pearson correlation coefficient fails the significance verification, it indicates that it may be due to random factors and lacks actual correlation significance, so it is removed from the initial correlation coefficient matrix. The correlation coefficients that pass the significance verification are retained, ultimately generating a correlation coefficient matrix between indicators and factors. This matrix contains statistically significant correlation coefficients that accurately reflect the true correlation between candidate remediation indicators and environmental factors.

[0048] Step 125: Based on the correlation coefficient matrix, target restoration indicators that meet the correlation requirements are selected. The selected target restoration indicators for vegetation growth status, soil physicochemical properties, and slope structural stability are integrated within the dimensions to generate a set of evaluation indicators for each dimension.

[0049] The correlation requirement can be set according to actual conditions. For example, a threshold for the correlation coefficient can be set, and candidate restoration indicators with correlation coefficients greater than this threshold can be selected as target restoration indicators. Target restoration indicators are selected for each of the following dimensions: vegetation growth status, soil physicochemical properties, and slope structural stability. Then, the selected target restoration indicators are integrated within each dimension. In the vegetation growth status dimension, the interrelationships between various target restoration indicators are analyzed. For example, there may be a correlation between vegetation coverage and vegetation biomass. By integrating these, a more comprehensive set of evaluation indicators that better reflects the vegetation growth status can be obtained. A similar method is used for correlation integration in the soil physicochemical properties and slope structural stability dimensions. The final set of evaluation indicators for each dimension accurately reflects the ecological restoration status of that dimension.

[0050] Step 126: Input the evaluation index sets of each dimension into the multi-dimensional association modeling strategy, establish the synergistic relationship between the evaluation index sets of each dimension through cross-dimensional interactive feature extraction processing, and construct a multi-dimensional evaluation model including vegetation growth status dimension, soil physicochemical property dimension and slope structural stability dimension based on the synergistic relationship.

[0051] After obtaining the evaluation index sets for each dimension, they need to be integrated to construct a multi-dimensional evaluation model. The evaluation index sets for each dimension are input into a multi-dimensional correlation modeling strategy, which aims to uncover the synergistic relationships between the dimensions. Cross-dimensional interactive feature extraction processing involves in-depth analysis of the evaluation index sets for each dimension to identify their mutual influences and correlations. For example, the evaluation index for the vegetation growth state dimension may be affected by the soil physicochemical properties dimension and the slope structural stability dimension. Through cross-dimensional interactive feature extraction processing, these potential influence relationships can be discovered. Based on these synergistic relationships, a multi-dimensional evaluation model is constructed that includes the vegetation growth state dimension, soil physicochemical properties dimension, and slope structural stability dimension. This model can comprehensively consider the interactions between multiple dimensions, providing a more comprehensive and accurate evaluation of the ecological restoration effect of steep slopes.

[0052] In an optional embodiment, step 126 includes: Step 1261: Input the evaluation index set of vegetation growth status dimension, the evaluation index set of soil physicochemical properties dimension, and the evaluation index set of slope structural stability dimension into the dimension feature layer of the multi-dimensional correlation modeling strategy.

[0053] The dimensional feature layer of the multi-dimensional correlation modeling strategy serves as the model's input layer, responsible for receiving evaluation indicator sets from each dimension. Evaluation indicator sets for vegetation growth status, soil physicochemical properties, and slope structural stability are input into the dimensional feature layer. The dimensional feature layer performs preliminary processing and analysis on these input evaluation indicator sets. For example, it checks the data format and integrity of each evaluation indicator set to ensure correct data processing. Simultaneously, the dimensional feature layer distinguishes and identifies evaluation indicator sets from different dimensions.

[0054] Step 1262: Process the evaluation index set of each dimension through the dimensional feature layer, and convert the index values ​​of different dimensions into the unprocessed index values ​​with a uniform distribution range to obtain the unprocessed evaluation index set of each dimension.

[0055] Because the evaluation index values ​​in each dimension may have different dimensions—for example, vegetation coverage in the vegetation growth status dimension is a percentage value, while soil moisture in the soil physicochemical properties dimension may be a specific moisture value—these different dimensions of index values ​​can cause difficulties in subsequent analysis and processing. Therefore, it is necessary to process the evaluation index sets of each dimension through a dimensional feature layer, converting the index values ​​with different dimensions into unprocessed index values ​​with a uniform distribution range. This conversion can be achieved using standardization or normalization methods. For example, for the index values ​​in each evaluation index set, calculate their mean and standard deviation, then subtract the mean from each index value and divide by the standard deviation to convert it into a standard normal distribution value. In this way, all index values ​​are converted to a uniform distribution range, resulting in the unprocessed evaluation index sets for each dimension, where the index values ​​have the same dimension and distribution range.

[0056] Step 1263: Perform principal component analysis on the set of evaluation indicators to be processed in each dimension, and extract the principal component feature vector of each dimension. The principal component feature vector is used to represent the core information of the dimension.

[0057] Optionally, principal component analysis (PCA) can be performed on the set of evaluation indicators to be processed in each dimension to extract the principal component eigenvectors for each dimension. PCA linearly combines multiple indicator values ​​in the set of evaluation indicators to generate a new set of variables, i.e., principal components. These principal components are mutually orthogonal and ordered according to their variance. The first few principal components typically contain most of the information from the original data. Through PCA, the principal component eigenvectors for each dimension are extracted. These eigenvectors represent the core information of that dimension. For example, in the vegetation growth status dimension, the principal component eigenvectors may contain comprehensive information on the main characteristics of vegetation growth, such as vegetation cover and vegetation biomass. By extracting the principal component eigenvectors, the high-dimensional set of evaluation indicators can be reduced to low-dimensional eigenvectors, reducing data complexity while retaining key information.

[0058] Step 1264: Input the principal component feature vectors of each dimension into the cross-dimensional interaction layer, extract the interaction features between dimensions by constructing a three-dimensional tensor product, and generate an interaction feature matrix containing the synergistic relationship between dimensions.

[0059] This method employs a three-dimensional tensor product (3D tensor product) to extract inter-dimensional interaction features. The 3D tensor product is a method that combines multiple vectors to capture the interactions between different dimensions. For the principal component feature vectors of the vegetation growth state dimension, soil physicochemical properties dimension, and slope structural stability dimension, a new interaction feature matrix is ​​generated through 3D tensor product operations. Each element in the matrix represents the interaction information between different dimensions. For example, an element in the matrix might represent the degree of synergy between the vegetation growth state dimension and the soil physicochemical properties dimension. The interaction feature matrix generated in this way contains the synergistic relationships between each dimension.

[0060] Step 1265: Perform feature selection processing on the interaction feature matrix, retain the interaction features whose variance contribution rate exceeds the preset contribution rate, and generate a key interaction feature set.

[0061] The interaction feature matrix may contain a large number of interaction features, but not all of them are significant for evaluating the ecological restoration effect of steep slopes. Therefore, feature selection processing is required for the interaction feature matrix. Feature selection processing determines the importance of each interaction feature by calculating its variance contribution rate. The variance contribution rate represents the degree to which the interaction feature contributes to the total variance. The preset contribution rate can be set according to actual conditions, for example, it can be set to 80%. Interaction features with variance contribution rates exceeding the preset contribution rate are retained to generate a key interaction feature set. The interaction features in this set are the key features that best reflect the synergistic relationship between the dimensions, which can effectively reduce data redundancy and improve the efficiency and accuracy of the model.

[0062] Step 1266: Concatenate the key interaction feature set with the principal component feature vectors of each dimension to generate a multi-dimensional comprehensive feature vector.

[0063] The concatenation process combines key interactive feature sets and principal component feature vectors in a specific order to form a new vector, which is the multi-dimensional comprehensive feature vector. This multi-dimensional comprehensive feature vector contains the core information of each dimension as well as the synergistic relationships between them. For example, it includes not only the core information of vegetation growth represented by the principal component feature vector of the vegetation growth state dimension, but also the synergistic information between the vegetation growth state dimension and the soil physicochemical properties and slope structural stability dimensions. Through this concatenation process, information from multiple dimensions is integrated into a single vector, providing a comprehensive and integrated input feature for constructing a multi-dimensional evaluation model.

[0064] Step 1267: Construct the node connection relationships of the input layer, hidden layer and output layer of the multi-dimensional evaluation model based on the multi-dimensional comprehensive feature vector, and generate the multi-dimensional evaluation model based on the node connection relationships.

[0065] In this model, the number of nodes in the input layer is the same as the dimension of the multi-dimensional comprehensive feature vector, and each node corresponds to an element in the multi-dimensional comprehensive feature vector. Multiple hidden layers can be set, each containing a certain number of nodes connected through a defined connection method, such as a fully connected layer. The function of the hidden layers is to perform non-linear transformations on the input multi-dimensional comprehensive feature vector, uncovering latent patterns and features in the data. The number of nodes in the output layer is set according to the evaluation requirements. For example, to evaluate the level of ecological restoration effectiveness on steep slopes, the number of nodes in the output layer can be set to the number of levels. Based on the above node connection relationships, a multi-dimensional evaluation model is generated. This model can output the evaluation results of the ecological restoration effectiveness on steep slopes based on the input multi-dimensional comprehensive feature vector.

[0066] Step 130: Use a time series analysis algorithm to track the dynamic trends of vegetation coverage, soil shear strength and slope erosion rate in the multi-dimensional evaluation model over a continuous period of time, generate index change tracking information, and combine the fuzzy comprehensive evaluation method to quantify the restoration effect level of the index change tracking information to obtain a comprehensive evaluation result of the ecological restoration effect of high and steep slopes.

[0067] To comprehensively understand the dynamic process and effects of ecological restoration of steep slopes, it is necessary to track the dynamic trends of key indicators in a multi-dimensional evaluation model over continuous time periods. Vegetation cover reflects the growth and coverage of vegetation on steep slopes, soil shear strength reflects the soil's ability to resist shear forces, and slope erosion rate indicates the speed at which the soil on the slope surface is eroded. Using time-series analysis algorithms to track these indicators can capture their changing trends over continuous time periods. The generated indicator change tracking information includes the trend direction, rate of change, and key inflection points of the indicators. The fuzzy comprehensive evaluation method is then used to quantify the restoration effect level of the indicator change tracking information. The fuzzy comprehensive evaluation method can handle uncertain and fuzzy information, matching the indicator change tracking information with preset restoration effect levels. Through a series of calculations and analyses, a comprehensive evaluation result of the ecological restoration effect of steep slopes is obtained. This result can provide decision-makers with intuitive and accurate information, helping them understand the actual situation of ecological restoration of steep slopes.

[0068] In one embodiment, step 130 includes: Step 131: Extract time-series data sequences of vegetation cover index, soil shear strength index and slope erosion rate index from the multi-dimensional evaluation model. The time-series data sequences contain index observations over consecutive time periods.

[0069] The multi-dimensional evaluation model records the observed values ​​of vegetation cover, soil shear strength, and slope erosion rate at different time points. Time-series data sequences for these indicators are extracted from the model; these sequences are arranged chronologically, with each data point corresponding to an observed value for a given time period. For example, for vegetation cover, the time-series data sequence might include monthly vegetation cover observations from the start of restoration to the present time. These time-series data sequences form the basis for subsequent time-series analysis, allowing us to understand the changes in these indicators over continuous time periods.

[0070] Step 132: Perform time series decomposition processing on the time series data sequence to separate the trend term, periodic term and random disturbance term. The trend term is used to represent the long-term change direction of the indicator, the periodic term is used to represent the seasonal fluctuation characteristics of the indicator, and the random disturbance term is used to represent the fluctuations caused by irregular factors.

[0071] Time series decomposition involves breaking down a time series data into three components: a trend term, a periodic term, and a random disturbance term. The trend term reflects the long-term direction of an indicator's change. For example, vegetation cover may gradually increase over time; this long-term increase is represented by the trend term. The periodic term reflects the seasonal fluctuations of an indicator. Some indicators may be affected by seasonal factors; for instance, vegetation cover may increase in spring due to vigorous growth and decrease in autumn. This seasonal fluctuation is represented by the periodic term. The random disturbance term represents fluctuations caused by irregular factors, such as sudden natural disasters or human interference. These fluctuations are random and unpredictable. Time series decomposition provides a clearer understanding of the changing characteristics of indicators.

[0072] Step 133: The trend term is smoothed using the sliding window averaging method to eliminate the influence of random fluctuations on trend judgment and obtain a smoothed trend sequence.

[0073] The trend item may still contain some random fluctuations, which can affect the accurate judgment of the long-term trend of the indicator. A sliding window averaging method is used to smooth the trend item. The sliding window averaging method sets a fixed-size window on the time series data of the trend item, calculates the average value of the data within the window, and then moves the window forward sequentially, calculating the average value of each window to obtain a new series, i.e., the smoothed trend series. This method can eliminate the influence of random fluctuations on trend judgment, making the trend clearer and more stable. For example, for the trend item of vegetation cover, some accidental factors may cause small fluctuations in the value at a certain point in time. The sliding window averaging method can smooth out these small fluctuations, more accurately showing the long-term trend of vegetation cover.

[0074] Step 134: Perform first-order difference processing on the smoothed trend sequence, calculate the index change at adjacent time points, and generate an index change rate sequence.

[0075] First-order differencing calculates the change in an indicator at adjacent time points within a smoothed trend sequence. For each data point in the smoothed trend sequence, the change in the indicator at adjacent time points is obtained by subtracting the data from the data at the previous time point from the data at the next time point. These changes are then combined into a new sequence, namely the indicator change rate sequence. The indicator change rate sequence reflects the rate of change of the indicator at each time point. For example, for a smoothed trend sequence of vegetation cover, calculating the change in vegetation cover between adjacent months yields an indicator change rate sequence that can show the rate of increase or decrease in vegetation cover each month.

[0076] Step 135: Based on the sign change pattern of the indicator change rate sequence, identify the turning point when the indicator changes from growth to decline or from decline to growth, and record the timestamp and indicator value corresponding to the turning point as turning point information.

[0077] The pattern of sign changes in an indicator's rate of change sequence reflects its growth or decline. When an element in the rate of change sequence changes from positive to negative, it indicates a shift from growth to decline; conversely, a change from negative to positive indicates a shift from decline to growth. Based on this pattern, inflection points can be identified. The timestamp and indicator value corresponding to each inflection point are recorded as inflection point information. For example, for the vegetation cover indicator, when an element in the rate of change sequence changes from positive to negative, the timestamp at that point and the corresponding vegetation cover value are recorded.

[0078] Step 136: Merge the smoothed trend sequence, the indicator change rate sequence, and the turning point information to generate indicator change tracking information that includes trend direction, change rate, and key turning points.

[0079] The smoothed trend series, indicator change rate series, and inflection point information are fused together. This fusion can be achieved by combining the above information according to a specific format to form a new dataset. This dataset serves as the indicator change tracking information, containing crucial information such as the indicator's trend direction, rate of change, and key inflection points. For example, the smoothed trend series reveals whether the indicator's long-term trend is increasing or decreasing; the indicator change rate series shows the rate of change at each time point; and the inflection point information identifies the key time points where the indicator underwent significant changes. By combining all this information, a comprehensive picture of the dynamic changes in vegetation cover, soil shear strength, and slope erosion rate indicators over a continuous period can be obtained.

[0080] In another embodiment, the step of combining the fuzzy comprehensive evaluation method to quantify the restoration effect level of the indicator change tracking information to obtain a comprehensive evaluation result of the ecological restoration effect of steep slopes includes: Step 1371: Determine the trend type of vegetation cover index, soil shear strength index and slope erosion rate index based on the trend direction in the index change tracking information. The trend type includes continuous growth, continuous decline, fluctuating growth and fluctuating decline.

[0081] The trend types of vegetation cover, soil shear strength, and slope erosion rate indices are determined based on the trend direction in the indicator change tracking information. For each indicator, if its trend direction is consistently upward without any significant decline over the entire time period, its trend type is determined as continuous growth; if its trend direction is consistently downward without any significant increase, it is determined as continuous decline; if the trend direction is generally upward but with some small downward fluctuations, it is determined as fluctuating growth; if the trend direction is generally downward but with some small upward fluctuations, it is determined as fluctuating decline. For example, for the vegetation cover indicator, if its smoothed trend sequence shows a continuous upward trend from the start of restoration to the present time without any significant downward phase, its trend type can be determined as continuous growth. By determining the above trend types, the changing characteristics of each indicator can be more clearly understood.

[0082] In a preferred embodiment, determining the change trend type of the vegetation cover index, soil shear strength index, and slope erosion rate index based on the trend direction in the index change tracking information includes: Step 13711: Extract the smoothed trend sequence from the indicator change tracking information, calculate the slope value of the trend sequence, and use the positive or negative value of the slope value to determine the trend direction.

[0083] The smoothed trend sequence is extracted from the indicator change tracking information, and the slope of this sequence is calculated. The slope value reflects the rate of change of the sequence at each time point, and its positive or negative value indicates the direction of the trend. If the slope value is positive, it means that the indicator is rising during that time period; if the slope value is negative, it means that the indicator is falling. For example, for the smoothed trend sequence of vegetation cover, the slope value at each time point is obtained by calculating the ratio of the difference between adjacent time points to the time interval. If most slope values ​​are positive, it means that vegetation cover generally shows an upward trend. By calculating the slope value, the trend direction of the indicator can be determined more accurately.

[0084] Step 13712: Calculate the proportion of the time length in which a positive slope appears consecutively in the trend sequence to the total time period. If the proportion exceeds a preset value, it is initially judged to be an upward trend.

[0085] The percentage of consecutive positive slopes in a smoothed trend sequence is used to determine the overall trend of the indicator. A preset ratio can be set based on specific circumstances, for example, 70%. If this ratio exceeds the preset value, it indicates that the indicator is rising for most of the time, and is initially judged as an upward trend. For example, for a smoothed trend sequence of vegetation cover, the number of months with consecutive positive slopes is counted, and then the percentage of this number to the total number of months is calculated. If this percentage exceeds 70%, the vegetation cover trend is initially judged to be an upward trend. This method can determine the overall trend of the indicator.

[0086] Step 13713: Calculate the proportion of time length in the trend sequence where negative slopes occur consecutively to the total time period. If the proportion exceeds a preset value, it is initially judged to be a downward trend.

[0087] Similarly, the proportion of consecutive periods with negative slopes in the smoothed trend sequence is used to determine the trend. When this proportion exceeds a preset value, it indicates that the indicator is decreasing for most of the time, and is initially judged as a downward trend. For example, for the smoothed trend sequence of slope erosion rate, the length of consecutive periods with negative slopes is counted, and the proportion of these periods to the total time is calculated. If this proportion exceeds a preset value, such as 70%, the trend of slope erosion rate is initially judged as a downward trend. In this way, a preliminary judgment of the downward trend of the indicator can be made, providing a reference for subsequent accurate classification.

[0088] Step 13714: For a sequence initially judged to be an upward trend, calculate its fluctuation range, which is the ratio of the difference between the maximum and minimum values ​​in the sequence to the mean of the sequence.

[0089] For sequences initially identified as having an upward trend, their volatility is further calculated. Volatility is an indicator that measures the degree of volatility in a sequence, and it is obtained by calculating the ratio of the difference between the maximum and minimum values ​​in the sequence to the sequence mean. For example, for a smoothed trend sequence of vegetation cover, first find the maximum and minimum values ​​in the sequence, calculate their difference, and then divide it by the sequence mean to obtain the volatility. Volatility reflects the fluctuation of the indicator during the growth process, providing an important basis for determining whether it is continuous growth or fluctuating growth.

[0090] Step 13715: If the fluctuation amplitude is less than the preset amplitude, the change trend type is determined to be continuous growth; if the fluctuation amplitude is greater than or equal to the preset amplitude, the change trend type is determined to be fluctuating growth.

[0091] The preset amplitude can be set according to actual conditions, for example, it can be set to 20%. The calculated fluctuation amplitude is compared with the preset amplitude. If the fluctuation amplitude is less than the preset amplitude, it means that the indicator fluctuates less and is relatively stable during the growth process, and the trend type is determined to be continuous growth; if the fluctuation amplitude is greater than or equal to the preset amplitude, it means that the indicator fluctuates more during the growth process, and the trend type is determined to be fluctuating growth. For example, for the smoothed trend sequence of vegetation cover, the calculated fluctuation amplitude is 15%, which is less than the preset amplitude of 20%, so its trend type is determined to be continuous growth. In this way, growth trends can be accurately classified.

[0092] Step 13716: Calculate the fluctuation amplitude of the sequence initially judged to be a downward trend. If the fluctuation amplitude is less than a preset amplitude, the trend type is determined to be a continuous downward trend; if the fluctuation amplitude is greater than or equal to the preset amplitude, the trend type is determined to be a fluctuating downward trend.

[0093] For sequences initially identified as having a downward trend, their fluctuation amplitude is also calculated. This amplitude is compared to a preset amplitude. If the amplitude is less than the preset amplitude, it indicates that the indicator is relatively stable with minimal fluctuations during the decline, and the trend type is determined to be a continuous decline. If the amplitude is greater than or equal to the preset amplitude, it indicates that the indicator exhibits significant fluctuations during the decline, and the trend type is determined to be a fluctuating decline. For example, for a smoothed trend sequence of slope erosion rates, if the calculated amplitude is 25%, which is greater than the preset amplitude of 20%, then the trend type is determined to be a fluctuating decline. This method allows for accurate classification of downward trends.

[0094] Step 1372: Based on the change trend type, set a fuzzy evaluation factor set for each indicator. The fuzzy evaluation factor set includes the trend stability, rate of change uniformity, and frequency of inflection point occurrence of the indicator.

[0095] Based on the determined trend types of vegetation cover, soil shear strength, and slope erosion rate indices, a fuzzy evaluation factor set is established for each index. This set includes three factors: trend stability, rate of change uniformity, and frequency of inflection points. Trend stability reflects the stability of the index during its change process; for example, continuously increasing indices have higher trend stability, while fluctuating indices have relatively lower trend stability. Rate of change uniformity refers to the consistency of the rate of change of the index during its change process; indices with uniform rates of change may be better in evaluation. Frequency of inflection points refers to the number of times an index changes from growth to decline or from decline to growth; indices with a lower frequency of inflection points may be more stable. The performance of these factors may differ for indices with different trend types. For example, for continuously increasing vegetation cover, its trend stability is high, its rate of change may be relatively uniform, and its frequency of inflection points is low; while for fluctuating vegetation cover, its trend stability is relatively low, its rate of change may be less uniform, and its frequency of inflection points may be high. By establishing a fuzzy evaluation factor set, the changes in the indices can be evaluated from multiple perspectives.

[0096] Step 1373: Construct a fuzzy evaluation level set for each indicator, wherein the fuzzy evaluation level set contains different levels of description of the repair effect.

[0097] A fuzzy evaluation level set is constructed for each indicator, containing different levels of description of restoration effectiveness. The number and specific descriptions of the levels can be set according to the actual situation; for example, four levels can be set: Excellent, Good, Average, and Poor. For the vegetation cover indicator, "Excellent" may indicate that vegetation cover has continued to grow rapidly, approaching or reaching the ideal coverage level; "Good" indicates that vegetation cover has increased to some extent, but with some small fluctuations; "Average" indicates that vegetation cover has grown slowly or fluctuated significantly; and "Poor" indicates that vegetation cover has not increased or has even decreased. For the soil shear strength and slope erosion rate indicators, corresponding level descriptions can also be set according to their changes and their impact on ecological restoration. The fuzzy evaluation level set provides clear evaluation criteria for subsequent fuzzy comprehensive evaluation. By matching the changes of the indicators with the above levels, the restoration effect can be quantified.

[0098] Step 1374: Based on the correspondence between the fuzzy evaluation factor set and the fuzzy evaluation level set, establish a fuzzy relation matrix for each indicator. The fuzzy relation matrix is ​​used to represent the membership degree of each factor in the factor set to each level in the level set.

[0099] Based on the correspondence between the fuzzy evaluation factor set and the fuzzy evaluation level set, a fuzzy relation matrix is ​​established for each indicator. Membership degree represents the degree of conformity between each factor in the factor set and each level in the level set. For example, for the trend stability factor of vegetation cover, its membership degree to the "excellent" level can be determined based on the specific manifestation of trend stability. If vegetation cover continues to increase and the trend is very stable, then the membership degree of the trend stability factor to the "excellent" level might be 0.8; to the "good" level, it might be 0.2; and to the "medium" and "poor" levels, it might be 0. By performing such judgments and calculations for each factor and each level, a matrix is ​​obtained, namely the fuzzy relation matrix. The rows of the matrix represent the factors in the fuzzy evaluation factor set, the columns represent the levels in the fuzzy evaluation level set, and each element in the matrix is ​​the membership degree of the corresponding factor to the level. By establishing the fuzzy relation matrix, the changes in indicators can be quantitatively correlated with the level of restoration effect.

[0100] Step 1375: Determine the weight distribution ratio of vegetation cover index, soil shear strength index and slope erosion rate index in the comprehensive evaluation using the analytic hierarchy process (AHP).

[0101] The Analytic Hierarchy Process (AHP) is a method used to determine the weight allocation of multiple factors in a comprehensive evaluation. For vegetation cover, soil shear strength, and slope erosion rate, the AHP is used to determine their weight allocation in the comprehensive evaluation. First, a hierarchical model is constructed, with the ecological restoration effect of steep slopes as the target layer and the three indicators as the criterion layer. Then, the indicators in the criterion layer are compared pairwise to determine their relative importance. For example, comparing the importance of vegetation cover and soil shear strength, if vegetation cover is considered more important in evaluating the ecological restoration effect, it can be assigned a higher relative importance value. Through a series of comparisons and calculations, the weight allocation ratio of each indicator is obtained. For example, the weight of vegetation cover might be 0.5, soil shear strength 0.3, and slope erosion rate 0.2. In this way, the importance of each indicator in the comprehensive evaluation can be reasonably determined.

[0102] Step 1376: Perform fuzzy matrix synthesis operation on the fuzzy relation matrix and the weight allocation ratio to generate fuzzy comprehensive evaluation vectors for each index.

[0103] The fuzzy relation matrix of each indicator is combined with its corresponding weight allocation ratio using a fuzzy matrix synthesis operation. Fuzzy matrix synthesis is a method that combines fuzzy relation matrices and weight vectors. For each indicator, its fuzzy relation matrix is ​​processed with the weight allocation ratio to obtain a new vector, namely the fuzzy comprehensive evaluation vector for each indicator. Each element in this vector represents the comprehensive membership degree of the indicator to each level in the fuzzy evaluation level set. For example, for the vegetation cover indicator, its fuzzy relation matrix is ​​processed with a weight of 0.5 to obtain a vector containing four elements, representing the comprehensive membership degree of the vegetation cover indicator to the four levels of "Excellent," "Good," "Medium," and "Poor." In this way, the evaluation results of multiple factors for each indicator are combined to obtain a more comprehensive evaluation vector.

[0104] Step 1377: Perform weighted averaging on the fuzzy comprehensive evaluation vectors of the aforementioned indicators to obtain the comprehensive evaluation vector of the ecological restoration effect of steep slopes.

[0105] A weighted average is applied to the fuzzy comprehensive evaluation vectors of each indicator. The fuzzy comprehensive evaluation vector for each indicator is multiplied by its corresponding weight allocation ratio, and then the results are summed to obtain the comprehensive evaluation vector for the ecological restoration effect of steep slopes. For example, the fuzzy comprehensive evaluation vector for vegetation cover is multiplied by a weight of 0.5, the fuzzy comprehensive evaluation vector for soil shear strength is multiplied by a weight of 0.3, and the fuzzy comprehensive evaluation vector for slope erosion rate is multiplied by a weight of 0.2. These three results are then summed to obtain a new vector, the comprehensive evaluation vector. Each element in this vector represents the comprehensive membership degree of the ecological restoration effect of steep slopes to each level in the fuzzy evaluation level set. By applying a weighted average, the evaluation results of the three indicators are combined to obtain a more comprehensive and accurate evaluation vector for the ecological restoration effect of steep slopes.

[0106] Step 1378: Traverse the membership values ​​of each level in the comprehensive evaluation vector, and determine the comprehensive evaluation result of the ecological restoration effect of the steep slope based on the level with the largest membership value; By iterating through the membership values ​​of each level in the comprehensive evaluation vector, the level with the largest membership value is identified. This level represents the comprehensive evaluation result of the ecological restoration effect on steep slopes. For example, if the comprehensive evaluation vector is [0.2, 0.5, 0.2, 0.1], representing the membership values ​​for the four levels "Excellent," "Good," "Medium," and "Poor," respectively, and the largest membership value is 0.5, corresponding to the level "Good," then the comprehensive evaluation result for the ecological restoration effect on steep slopes is determined to be "Good." This method transforms the fuzzy evaluation result into a specific level, providing decision-makers with intuitive and clear information, facilitating their understanding of the actual effects of ecological restoration on steep slopes.

[0107] Optionally, the step of traversing the membership values ​​of each level in the comprehensive evaluation vector and determining the comprehensive evaluation result of the ecological restoration effect of steep slopes based on the level with the largest membership value includes: Step 13781: Extract the membership values ​​corresponding to each level in the comprehensive evaluation vector. The membership values ​​are used to indicate the degree to which the comprehensive evaluation result belongs to that level.

[0108] The membership values ​​corresponding to each level are extracted from the comprehensive evaluation vector. These membership values ​​reflect the degree to which the ecological restoration effect of steep slopes belongs to each level. For example, the comprehensive evaluation vector is [0.1, 0.6, 0.2, 0.1], corresponding to the membership values ​​of the four levels of "Excellent", "Good", "Medium", and "Poor", respectively. 0.6 indicates that the degree to which the ecological restoration effect of steep slopes belongs to the "Good" level is 60%. By extracting these membership values, we can better understand the probability of each level.

[0109] Step 13782: Compare the membership values ​​of each level, determine the level with the largest membership value, and use this level as the preliminary evaluation result.

[0110] Compare the membership values ​​of each level in the comprehensive evaluation vector and find the level corresponding to the largest membership value. This level is taken as the preliminary evaluation result. For example, in the comprehensive evaluation vector [0.1, 0.6, 0.2, 0.1], the largest membership value is 0.6, corresponding to the level "Good," so the preliminary evaluation result is "Good." This simple comparison method can quickly obtain a preliminary evaluation result, but its reliability still needs further verification.

[0111] Step 13783: Check the difference between the membership value corresponding to the preliminary evaluation result and the membership values ​​of other levels. If the difference is less than the preset difference, then perform a second evaluation.

[0112] Check the difference between the membership value corresponding to the preliminary evaluation result and the membership values ​​of other levels. The preset difference can be set according to the actual situation; for example, it can be set to 0.2. If the difference between the membership value corresponding to the preliminary evaluation result and the membership values ​​of other levels is less than the preset difference, it indicates that the distinction between the levels is not significant, the reliability of the preliminary evaluation result is low, and a second evaluation is required. This checking method can improve the accuracy and reliability of the evaluation results.

[0113] Step 13784: When conducting the secondary evaluation, calculate the weighted membership value for each level. The weighted membership value is the product of the membership value and the reference score for that level.

[0114] During the secondary evaluation, a weighted membership value is calculated for each level. The reference score can be set according to the actual situation; for example, for the four levels "Excellent," "Good," "Average," and "Poor," reference scores of 90, 70, 50, and 30 can be set respectively. The membership value for each level is multiplied by its reference score to obtain the weighted membership value.

[0115] Step 13785: Sum the weighted membership values ​​and divide by the total membership values ​​to obtain the comprehensive evaluation score.

[0116] The weighted membership values ​​of each level are summed, and then divided by the total membership values ​​to obtain the comprehensive evaluation score. This calculation method transforms the fuzzy evaluation results into a specific score, which can more intuitively reflect the quality of ecological restoration of steep slopes.

[0117] Step 13786: Determine the corresponding evaluation level based on the score range of the comprehensive evaluation score, and determine the comprehensive evaluation result of the ecological restoration effect of the steep slope based on the evaluation level.

[0118] The evaluation level is determined based on the score range of the comprehensive evaluation. Score ranges and corresponding level descriptions can be set according to the actual situation; for example, 90-100 points is "Excellent," 70-89 points is "Good," 50-69 points is "Average," and 30-49 points is "Poor." If the comprehensive evaluation score is 59 points, falling within the 50-69 point range, the corresponding evaluation level is "Average," and the comprehensive evaluation result for the ecological restoration effect of the steep slope is determined to be "Average." In this way, the comprehensive evaluation score is converted into a specific evaluation level, providing decision-makers with clear information to facilitate the development of appropriate restoration strategies based on the evaluation results.

[0119] Step 140: Based on the comprehensive evaluation results, generate a visual evaluation report including a time-axis evolution trend map, a spatial distribution heat map, and a boundary line for the classification of restoration levels, and push the visual evaluation report to the ecological restoration management platform to support the decision-making on the adjustment of restoration strategies.

[0120] In this embodiment, to more intuitively demonstrate the ecological restoration effect of steep slopes, a visual assessment report is generated based on the comprehensive evaluation results. This report includes a timeline evolution trend map, a spatial distribution heat map, and restoration level classification boundaries. These visualization elements can display the ecological restoration status of steep slopes from different perspectives. The timeline evolution trend map shows the changes in restoration effects over time, helping decision-makers understand the progress and trends of restoration; the spatial distribution heat map shows the distribution of restoration effects in different areas of the steep slope, identifying areas with better and worse restoration effects; and the restoration level classification boundaries clearly define the boundaries between areas of different restoration levels, facilitating targeted management. The visual assessment report is pushed to the ecological restoration management platform, where decision-makers can adjust restoration strategies based on the information in the report, such as adding restoration measures in areas with poor restoration effects and maintaining or optimizing existing measures in areas with better effects, thereby improving the overall effectiveness of steep slope ecological restoration.

[0121] In one design approach, based on the comprehensive evaluation results, a visual assessment report is generated that includes a timeline evolution trend map, a spatial distribution heat map, and boundary lines for the classification of repair levels. This includes: Step 141: Extract the repair effect level data for different time periods from the comprehensive evaluation results, associate the repair effect level data with the corresponding timestamps, and generate a timeline data sequence.

[0122] The restoration effect level data for different time periods are extracted from the comprehensive evaluation results. This data records the restoration effect levels of steep slopes at different times, such as "Excellent," "Good," "Medium," and "Poor." Each restoration effect level is associated with a corresponding timestamp; for example, the restoration effect level "Good" for a certain month is associated with the specific date of that month. In this way, a timeline data sequence arranged chronologically is generated. This sequence forms the basis for plotting the timeline evolution trend diagram, clearly showing how the restoration effect changes over time.

[0123] Step 142: Draw a timeline evolution trend chart based on the timeline data sequence, and connect the level data of each time point with a broken line; the horizontal axis of the timeline evolution trend chart is the timestamp, and the vertical axis is the repair effect level.

[0124] Using timestamps as the horizontal axis and restoration effectiveness levels as the vertical axis, each data point in the timeline data sequence is labeled on the graph. A line graph connects the level data at each time point, providing a visual representation of the changing trend of restoration effectiveness levels over time. For example, an upward trend indicates that the restoration effectiveness is gradually improving, while a downward trend indicates that the restoration effectiveness is deteriorating. This timeline evolution trend graph helps decision-makers quickly understand the long-term progress of ecological restoration of steep slopes.

[0125] Step 143: Extract the geographical coordinates of each sampling point and the corresponding restoration effect level data from the slope restoration observation dataset to generate a spatial location-level dataset.

[0126] The geographic coordinates and corresponding restoration effect level data of each sampling point are extracted from the previously generated slope restoration observation dataset. The geographic coordinates precisely record the specific location of each sampling point in the steep slope area, while the restoration effect level data reflects the ecological restoration effect at that sampling point. These two are correlated to generate a new dataset: a spatial location-level dataset. For example, for a given sampling point, its geographic coordinates are set as latitude and longitude, and its corresponding restoration effect level is "good." Combining these coordinates forms a data point in the dataset. This dataset contains the spatial location and restoration effect information of each sampling point within the steep slope area, serving as the basis for drawing spatial distribution heat maps.

[0127] Step 144: Based on the spatial location-level dataset, an interpolation algorithm is used to generate spatially continuous repair effect level distribution data. The interpolation algorithm is used to convert discrete sampling point data into continuous spatial distribution data.

[0128] Since the spatial location-grade dataset consists of discrete sampling point data, interpolation algorithms are needed to obtain a continuous distribution of repair effect grades across the entire steep slope area. The core idea of ​​interpolation algorithms is to infer the repair effect grade data for unknown locations based on known discrete sampling point data. For example, if there are several sampling points within a steep slope area, and their repair effect grades are known, the repair effect grade at a given location can be estimated using interpolation algorithms based on the grade data of adjacent sampling points. Common interpolation algorithms include Kriging interpolation and linear interpolation. By using interpolation algorithms, discrete sampling point data is transformed into continuous spatial distribution data, allowing for a more accurate description of the repair effect grade distribution across the entire steep slope area.

[0129] Step 145: Draw a spatial distribution heat map based on the spatial continuous repair effect level distribution data. The color intensity of the spatial distribution heat map is used to represent the high and low distribution of the repair effect level.

[0130] During the mapping process, different colors were assigned to different restoration effectiveness levels, with the shade of the color corresponding to the severity of the restoration. For example, areas rated "Excellent" were represented by a darker green; areas rated "Good" by a lighter green; areas rated "Medium" by yellow; and areas rated "Poor" by red. This color-coding method visually displays the distribution of restoration effectiveness levels for steep slopes in a heatmap format. Decision-makers can quickly understand the restoration effectiveness of different areas of steep slopes by observing the color distribution on the heatmap, identifying areas with better and worse restoration results.

[0131] Step 146: Perform region division processing on the repair effect level data in the spatial distribution heat map, identify connected regions with the same repair effect level, and extract the boundary contour lines of each connected region.

[0132] The restoration effect level data in the drawn spatial distribution heatmap is divided into regions. This process involves identifying connected regions with the same restoration effect level; that is, adjacent regions with the same restoration effect level are considered as a connected region. For example, in the heatmap, a continuous green area represents a connected region with a restoration effect level of "good". Then, the boundary contour lines of each connected region are extracted. Boundary contour line extraction can be achieved using techniques such as image recognition and edge detection. These boundary contour lines clearly define the boundaries between regions with different restoration effect levels, providing a basis for subsequent generation of restoration effect level division boundary lines. This helps decision-makers better understand the distribution range of different restoration effect areas on steep slopes.

[0133] Step 147: Overlay the boundary contour lines onto the spatial distribution heat map to generate repair level classification boundary lines.

[0134] The extracted boundary contours of each connected region are overlaid onto the spatial distribution heatmap to generate boundary lines for different restoration levels. The overlay process involves accurately placing the boundary contours at their corresponding positions on the heatmap, ensuring the boundary lines match the restoration effect level regions. This clearly shows the boundaries between different restoration effect levels on the spatial distribution heatmap. These restoration level boundary lines further enhance the visualization of the heatmap, allowing decision-makers to more intuitively distinguish regions of different restoration levels and providing clearer spatial information for developing precise restoration strategies.

[0135] Step 148: Integrate the time axis evolution trend map, spatial distribution heat map and repair level classification boundary line into the assessment report template to generate a visual assessment report containing chart titles, coordinate axis descriptions and legends.

[0136] The pre-drawn timeline evolution trend chart, spatial distribution heat map, and generated restoration level classification boundary lines are integrated into the assessment report template. The assessment report template is pre-designed and includes a framework of elements such as chart titles, axis descriptions, and legends. During integration, the timeline evolution trend chart is placed in an appropriate position and a chart title is added, such as "Timeline Evolution Trend Chart of Ecological Restoration Effects on Steep Slopes." Detailed explanations are provided for the horizontal and vertical axes, such as labeling the horizontal axis as "Time (Month)" and the vertical axis as "Restoration Effect Level." Similarly, a corresponding chart title is added to the spatial distribution heat map, such as "Spatial Distribution Heat Map of Ecological Restoration Effects on Steep Slopes," and legends are provided to explain the restoration effect levels represented by colors, such as "Green - Excellent, Light Green - Good, Yellow - Medium, Red - Poor." The restoration level classification boundary lines are accurately overlaid on the spatial distribution heat map and explained in the report. In this way, a complete and information-rich visual assessment report is generated, which comprehensively and intuitively displays the temporal and spatial changes in the ecological restoration effects of steep slopes.

[0137] As a non-limiting embodiment, the method further includes: acquiring a slope restoration observation dataset for a continuous time period after the visualization assessment report is pushed; using the same time-series analysis algorithm as in the multi-dimensional evaluation model to dynamically track the trends of vegetation cover, soil shear strength, and slope erosion rate in the slope restoration observation dataset for the continuous time period, generating the latest indicator change tracking information; comparing the latest indicator change tracking information with the indicator change tracking information to generate a trend consistency determination result; if the trend consistency determination result indicates that the trends are consistent, then outputting a trend continuity report; if the trend consistency determination result indicates that the trends are inconsistent, then extracting abnormal fluctuation features from the latest indicator change tracking information, generating a trend anomaly analysis report.

[0138] After pushing the visualization assessment report to the ecological restoration management platform, it is necessary to continuously monitor the ecological restoration status of steep slopes. Obtain the slope restoration observation dataset for consecutive time periods following the report push. This dataset contains the same types of data as before, such as land cover spectral data, soil environmental monitoring data, slope surface image data, and vegetation physiological characteristic data. Using the same time-series analysis algorithm as in the multi-dimensional evaluation model, dynamically track the trends of vegetation cover, soil shear strength, and slope erosion rate indices in the continuous time-series slope restoration observation dataset. The specific process is similar to the previous generation of index change tracking information, including steps such as extracting time-series data sequences, performing time-series decomposition, smoothing, first-order differencing, and identifying inflection points, ultimately generating the latest index change tracking information.

[0139] The latest indicator change tracking information is compared with previously generated indicator change tracking information to ensure trend consistency. This comparison involves examining the trend direction, rate of change, and key inflection points of the indicators in the two tracking sets. For example, it checks whether the vegetation cover indicator shows an increasing trend in both tracking sets, and whether the rates of change are similar. Based on the comparison results, a trend consistency judgment is generated, which is categorized into two cases: consistent trend and inconsistent trend.

[0140] If the trend consistency assessment indicates that the trend is consistent, it means that the ecological restoration of steep slopes is continuing according to the previous trend, and a trend continuity report is generated. The report details the continuous trend of each indicator, such as the continuous increase in vegetation cover and the continuous enhancement of soil shear strength, providing decision-makers with information on the stable development of restoration effects so that they can decide whether to maintain the existing restoration strategy.

[0141] If the trend consistency assessment indicates inconsistency, it suggests a change in the ecological restoration status of steep slopes, necessitating the extraction of anomalous fluctuation characteristics from the latest indicator tracking information. Anomalous fluctuation characteristic extraction analyzes in which aspects of the indicators have deviated from previous levels, such as a sudden decrease in vegetation cover or a sudden increase in slope erosion rate. Based on the extracted anomalous fluctuation characteristics, a trend anomaly analysis report is generated. This report details the circumstances and possible causes of the anomalous fluctuations, providing a basis for decision-makers to adjust restoration strategies to address anomalies encountered during ecological restoration.

[0142] As a non-limiting embodiment, the method further includes: acquiring a slope repair observation dataset for a continuous time period after the repair strategy adjustment; performing quantification of the repair effect level on the slope repair observation dataset for the continuous time period after the repair strategy adjustment based on the multi-dimensional evaluation model to generate a comprehensive evaluation result after strategy adjustment; comparing the comprehensive evaluation result after strategy adjustment with the comprehensive evaluation result to calculate a model prediction error sequence; performing correlation feature contribution analysis on the correlation mapping relationship in the multi-dimensional evaluation model based on the model prediction error sequence to generate a feature contribution change rate; filtering out sensitive correlation features whose contribution changes exceed a preset threshold based on the feature contribution change rate; updating the feature weight coefficients of the correlation mapping relationship corresponding to the sensitive correlation features to generate an updated correlation feature weight coefficient set; and adjusting the parameters of the multi-dimensional evaluation model using the updated correlation feature weight coefficient set to obtain an optimized multi-dimensional evaluation model.

[0143] After the ecological restoration management platform adjusts the restoration strategy based on the visualization assessment report, it obtains a continuous time period of slope restoration observation dataset after the adjustment of the restoration strategy. This dataset records the ecological restoration status of steep slopes under the new restoration strategy and includes various observation data.

[0144] Based on the previously constructed multi-dimensional evaluation model, the slope restoration observation dataset over a continuous time period after the restoration strategy adjustment was used to quantify the restoration effect level. The processing procedure was similar to that used to generate the comprehensive evaluation results, including generating the correlation mapping relationship between restoration indicators and environmental factors, constructing a multi-dimensional evaluation model, performing time series analysis and fuzzy comprehensive evaluation, and finally generating the comprehensive evaluation results after the strategy adjustment.

[0145] The comprehensive evaluation results after strategy adjustment are compared with the previous comprehensive evaluation results, and the difference between the two is calculated to obtain the model prediction error sequence. This sequence reflects the error of the multi-dimensional evaluation model in predicting the repair effect.

[0146] This study analyzes the contribution of correlation features in a multi-dimensional evaluation model based on the model prediction error sequence. The contribution analysis assesses the influence of each correlation feature on the model's prediction results. By analyzing the model prediction error sequence, the contribution of each correlation feature to the error changes is calculated, generating a feature contribution change rate. This rate represents the change in the contribution of each correlation feature before and after the adjustment of the repair strategy.

[0147] Sensitive correlation features are selected based on the rate of change in feature contribution, where the change exceeds a preset threshold. The preset threshold can be set according to actual conditions, for example, to 30%. Correlation features whose contribution changes exceed this threshold are considered sensitive correlation features, indicating that their impact on the model's prediction results has changed significantly after the adjustment of the repair strategy.

[0148] The feature weight coefficients of the correlation mapping relationships corresponding to sensitive correlation features are updated. The weight coefficients of these correlation features are adjusted according to the rate of change of feature contribution, enabling the model to more accurately reflect the actual situation after the adjustment of the repair strategy. The updated feature weight coefficients constitute the updated set of correlation feature weight coefficients.

[0149] The parameters of the multi-dimensional evaluation model are adjusted using an updated set of associated feature weight coefficients. The adjustment process involves applying the updated weight coefficients to the node connections in the model's input, hidden, and output layers, thus resetting the model's parameters. In this way, an optimized multi-dimensional evaluation model is obtained, which can more accurately evaluate the ecological restoration effects of steep slopes under the new restoration strategy.

[0150] As a non-limiting embodiment, the method further includes: obtaining a comprehensive evaluation result of the new ecological restoration effect of steep slopes; comparing and analyzing the new comprehensive evaluation result of the ecological restoration effect of steep slopes with the existing comprehensive evaluation result to generate an evaluation result difference index; determining whether the visualization assessment report needs to be updated based on the evaluation result difference index; if an update is required, updating the time axis evolution trend map, spatial distribution heat map, and restoration level division boundary line based on the new comprehensive evaluation result of the ecological restoration effect of steep slopes to generate an updated time axis evolution trend map, an updated spatial distribution heat map, and an updated restoration level division boundary line; integrating the updated time axis evolution trend map, the updated spatial distribution heat map, and the updated restoration level division boundary line to generate an updated visualization assessment report, and pushing the updated visualization assessment report to the ecological restoration management platform.

[0151] Understandably, the process involves comparing and analyzing the new comprehensive evaluation results with the previous ones to generate an evaluation result difference index. This comparative analysis involves comparing information such as the repair effectiveness level data and spatial distribution across different time periods in the two comprehensive evaluation results. For example, it examines whether the repair effectiveness level has changed at the same time point, and whether there are differences in the distribution of repair effectiveness levels in different regions. Based on the comparison results, the evaluation result difference index is calculated, which reflects the degree of difference between the two comprehensive evaluation results.

[0152] The need to update the visualization assessment report is determined based on the difference index of the evaluation results. A difference threshold can be set, such as 20%. If the difference index exceeds this threshold, it indicates a significant change in the ecological restoration of the steep slope, requiring an update to the visualization assessment report; if the difference index does not exceed the threshold, it indicates a minor change, and an update is not necessary at this time.

[0153] If an update is deemed necessary, the timeline evolution trend map, spatial distribution heatmap, and restoration level classification boundary lines are updated based on the comprehensive evaluation results of the new ecological restoration effects on steep slopes. For the timeline evolution trend map, the new restoration effect level data is added to the timeline data series, and the line graph is redrawn to show the latest changes in restoration effects over time. For the spatial distribution heatmap, interpolation processing is performed again and the heatmap is redrawn based on the restoration effect level data for each sampling point in the new comprehensive evaluation results, updating the color distribution. For the restoration level classification boundary lines, regional division processing and boundary contour line extraction are performed again to reflect the latest restoration level regional boundaries.

[0154] The updated timeline evolution trend map, the updated spatial distribution heat map, and the updated restoration level classification boundary line are integrated into the assessment report template to generate an updated visual assessment report. This report contains the latest information on the ecological restoration effect of steep slopes and is pushed to the ecological restoration management platform to provide decision-makers with the latest and most accurate information so that they can adjust restoration strategies in a timely manner and promote the continuous and effective implementation of ecological restoration work on steep slopes.

[0155] This application's embodiments achieve a comprehensive, dynamic, and accurate evaluation of the ecological restoration effect of steep slopes through a series of steps, including the integrated use of multi-source observation data, correlation mapping analysis, multi-dimensional evaluation model construction, time-series analysis and fuzzy evaluation, and visualization report generation. A slope restoration observation dataset with time-series correlation is formed using multi-source observation data from steep slope areas, integrating data from remote sensing satellites, sensor networks, UAV platforms, and field sampling, overcoming the limitations of incomplete information from a single data source. Based on this dataset, a correlation mapping relationship between restoration indicators and environmental factors is generated, and a multi-dimensional evaluation model is constructed. This model comprehensively considers multiple dimensions, including vegetation growth status, soil physicochemical properties, and slope structural stability, providing a more comprehensive and in-depth reflection of the actual situation of ecological restoration of steep slopes, avoiding the one-sidedness of traditional evaluation methods that only focus on a single dimension. The use of time-series analysis algorithms for dynamic trend tracking of key indicators, combined with fuzzy comprehensive evaluation methods for quantifying the restoration effect level, not only captures dynamic changes during the ecological restoration process but also effectively handles uncertainties and fuzziness in the evaluation, improving the accuracy and reliability of the evaluation results. Based on the comprehensive evaluation results, a visual assessment report is generated, which includes a timeline evolution trend map, a spatial distribution heat map, and boundary lines for the classification of restoration levels. This report is then pushed to the management platform, presenting the complex evaluation results in an intuitive chart format. This allows decision-makers to quickly understand the temporal and spatial variation characteristics of ecological restoration of steep slopes, thereby enabling them to adjust restoration strategies in a timely and accurate manner and improving the scientificity and effectiveness of ecological restoration management.

[0156] Based on the same inventive concept, this application also provides an ecological restoration evaluation system for steep slopes. (See also...) Figure 2 As shown, this is a schematic diagram of a possible ecological restoration evaluation system for steep slopes provided in an embodiment of this application. Figure 2 The steep slope ecological restoration evaluation system 200 includes a processor 210 and a memory 220. The memory 220 stores computer programs executable by the processor 210. By executing the instructions stored in the memory 220, the processor 210 can perform the steps of the aforementioned steep slope ecological restoration evaluation method based on multi-source data fusion.

[0157] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium including a computer program. When the computer program is run on a steep slope ecological restoration evaluation system, the computer program is used to cause the steep slope ecological restoration evaluation system to perform the steps of the aforementioned steep slope ecological restoration evaluation method based on multi-source data fusion. In some possible implementations, various aspects of the steep slope ecological restoration evaluation method based on multi-source data fusion provided in this application can also be implemented in the form of a program product, including a computer program. When the program product is run on a steep slope ecological restoration evaluation system, the computer program is used to cause the steep slope ecological restoration evaluation system to perform the steps of the aforementioned steep slope ecological restoration evaluation method based on multi-source data fusion. For example, the steep slope ecological restoration evaluation system can perform actions such as... Figure 1 The steps are shown in the figure.

Claims

1. A method for evaluating the ecological restoration of steep slopes based on multi-source data fusion, characterized in that, The method includes: Using multi-source observation data from high and steep slope areas, a slope restoration observation dataset with time-series correlation is formed; wherein, the slope restoration observation dataset includes land cover spectral data collected by remote sensing satellites, soil environmental monitoring data collected by deploying sensor networks, slope surface image data collected by UAV platforms, and vegetation physiological characteristic data obtained through field sampling. Based on the slope restoration observation dataset, a correlation mapping relationship between restoration indicators and environmental factors is generated. Based on the correlation mapping relationship, a multi-dimensional evaluation model including vegetation growth status, soil physicochemical properties, and slope structural stability is constructed. The time series analysis algorithm is used to track the dynamic trends of vegetation coverage, soil shear strength and slope erosion rate in the multi-dimensional evaluation model over a continuous period of time, generating index change tracking information. The fuzzy comprehensive evaluation method is then used to quantify the restoration effect level of the index change tracking information, and a comprehensive evaluation result of the ecological restoration effect of high and steep slopes is obtained. Based on the comprehensive evaluation results, a visual assessment report is generated, which includes a time-axis evolution trend map, a spatial distribution heat map, and boundary lines for the classification of restoration levels. The visual assessment report is then pushed to the ecological restoration management platform to support the decision-making on the adjustment of restoration strategies.

2. The method according to claim 1, characterized in that, The process involves generating a correlation mapping relationship between restoration indicators and environmental factors based on the slope restoration observation dataset, and constructing a multi-dimensional evaluation model based on this correlation mapping relationship, including dimensions of vegetation growth status, soil physicochemical properties, and slope structural stability. Extract the first observation variable related to vegetation growth status from the slope restoration observation dataset, and use the first observation variable as the first candidate restoration index for the vegetation growth status dimension. A second observation variable related to soil physicochemical properties is extracted from the slope restoration observation dataset, and the second observation variable is used as the second candidate restoration index for the soil physicochemical property dimension. A third observation variable related to slope structural stability is extracted from the slope restoration observation dataset, and the third observation variable is used as the third candidate restoration index for the slope structural stability dimension. Based on the statistical correlation between each candidate remediation index and its corresponding environmental factor, a correlation coefficient matrix between the index and the factor is generated through bivariate correlation analysis. Based on the correlation coefficient matrix, target restoration indicators that meet the correlation requirements are selected. The selected target restoration indicators of vegetation growth status, soil physicochemical properties and slope structural stability are integrated within the dimensions to generate a set of evaluation indicators for each dimension. The evaluation index sets of each dimension are input into the multi-dimensional correlation modeling strategy. The synergistic relationship between the evaluation index sets of each dimension is established through cross-dimensional interactive feature extraction processing. Based on the synergistic relationship, a multi-dimensional evaluation model including vegetation growth status dimension, soil physicochemical property dimension and slope structural stability dimension is constructed.

3. The method according to claim 2, characterized in that, The method, based on the statistical correlation between each candidate remediation indicator and its corresponding environmental factor, generates a correlation coefficient matrix between the indicator and the factor through bivariate correlation analysis, including: The time series data of each candidate remediation indicator is synchronized and matched with the corresponding environmental factor time series data to obtain a time series synchronization matching dataset; The time-series synchronization matching dataset is subjected to stationarity verification. If the data does not meet the stationarity requirements, differential processing is performed until the time-series synchronization matching dataset passes the stationarity verification. A bivariate correlation analysis was performed on the time-series synchronous matching dataset that passed the stationarity verification to obtain the Pearson correlation coefficient; wherein, the Pearson correlation coefficient is used to represent the degree of linear association between candidate remediation indicators and environmental factors; The Pearson correlation coefficients are arranged according to the cross-combination of candidate remediation indicators and environmental factors to generate a preliminary correlation coefficient matrix. The preliminary correlation coefficient matrix is ​​then subjected to a significance level check. Correlation coefficients that fail the significance check are removed, while those that pass the significance check are retained, thus generating a correlation coefficient matrix between indicators and factors.

4. The method according to claim 2, characterized in that, The process involves inputting evaluation index sets from various dimensions into a multi-dimensional correlation modeling strategy, establishing synergistic relationships between these evaluation index sets through cross-dimensional interactive feature extraction, and constructing a multi-dimensional evaluation model based on these synergistic relationships. This model includes dimensions of vegetation growth status, soil physicochemical properties, and slope structural stability. The evaluation index sets for vegetation growth status, soil physicochemical properties, and slope structural stability are input into the dimensional feature layer of the multi-dimensional correlation modeling strategy. The evaluation index set of each dimension is processed by the dimensional feature layer, and the index values ​​of different dimensions are converted into the index values ​​to be processed within a uniform distribution range, thus obtaining the evaluation index set to be processed for each dimension. Principal component analysis is performed on the set of evaluation indicators to be processed in each dimension to extract the principal component feature vector of each dimension. The principal component feature vector is used to represent the core information of that dimension. The principal component feature vectors of each dimension are input into the cross-dimensional interaction layer. Inter-dimensional interaction features are extracted by constructing a three-dimensional tensor product, generating an interaction feature matrix that contains the synergistic relationship between dimensions. The interaction feature matrix is ​​subjected to feature selection processing to retain the interaction features whose variance contribution rate exceeds the preset contribution rate, thereby generating a key interaction feature set. The key interactive feature set is concatenated with the principal component feature vectors of each dimension to generate a multi-dimensional comprehensive feature vector. Based on the multi-dimensional comprehensive feature vector, the node connection relationships of the input layer, hidden layer and output layer of the multi-dimensional evaluation model are constructed, and the multi-dimensional evaluation model is generated based on the node connection relationships.

5. The method according to claim 1, characterized in that, The method employs a time-series analysis algorithm to dynamically track the vegetation cover, soil shear strength, and slope erosion rate indices in the multi-dimensional evaluation model over a continuous time period, generating index change tracking information, including: The time-series data of vegetation cover index, soil shear strength index and slope erosion rate index are extracted from the multi-dimensional evaluation model. The time-series data includes the index observation values ​​for consecutive time periods. The time series data sequence is subjected to time series decomposition processing to separate the trend term, periodic term and random disturbance term. The trend term is used to represent the long-term change direction of the indicator, the periodic term is used to represent the seasonal fluctuation characteristics of the indicator, and the random disturbance term is used to represent the fluctuation caused by irregular factors. The trend term is smoothed using a sliding window averaging method to eliminate the influence of random fluctuations on trend judgment, resulting in a smoothed trend sequence. The smoothed trend sequence is subjected to first-order difference processing to calculate the index change at adjacent time points and generate an index change rate sequence. Based on the sign change pattern of the indicator change rate sequence, identify the turning point when the indicator changes from growth to decline or from decline to growth, and record the timestamp and indicator value corresponding to the turning point as turning point information. The smoothed trend sequence, the indicator change rate sequence, and the turning point information are fused together to generate indicator change tracking information that includes trend direction, change rate, and key turning points.

6. The method according to claim 5, characterized in that, The method of combining fuzzy comprehensive evaluation to quantify the restoration effect level of the indicator change tracking information yields a comprehensive evaluation result of the ecological restoration effect of steep slopes, including: Based on the trend direction in the indicator change tracking information, the change trend type of vegetation coverage index, soil shear strength index and slope erosion rate index is determined. The change trend type includes continuous growth, continuous decline, fluctuating growth and fluctuating decline. Based on the aforementioned trend type, a fuzzy evaluation factor set is set for each indicator, which includes the trend stability, rate of change uniformity, and frequency of inflection point occurrence of the indicator. Construct a fuzzy evaluation level set for each indicator, wherein the fuzzy evaluation level set contains different levels of description of the repair effect; Based on the correspondence between the fuzzy evaluation factor set and the fuzzy evaluation level set, a fuzzy relation matrix for each indicator is established. The fuzzy relation matrix is ​​used to represent the membership degree of each factor in the factor set to each level in the level set. The weighting ratios of vegetation cover, soil shear strength and slope erosion rate in the comprehensive evaluation were determined by the analytic hierarchy process. The fuzzy relation matrix and the weight allocation ratio are combined using a fuzzy matrix synthesis operation to generate a fuzzy comprehensive evaluation vector for each indicator. The fuzzy comprehensive evaluation vectors of the aforementioned indicators are weighted and averaged to obtain the comprehensive evaluation vector of the ecological restoration effect of steep slopes. The membership values ​​of each level in the comprehensive evaluation vector are traversed, and the comprehensive evaluation result of the ecological restoration effect of the steep slope is determined according to the level with the largest membership value. The determination of the trend types of vegetation cover index, soil shear strength index, and slope erosion rate index based on the trend direction in the index change tracking information includes: Extract the smoothed trend sequence from the indicator change tracking information, calculate the slope value of the trend sequence, and use the positive or negative value of the slope value to determine the trend direction; The proportion of consecutive positive slopes in the trend sequence is counted to the total time period. If the proportion exceeds a preset value, it is initially judged as an upward trend. The proportion of consecutive negative slopes in the trend sequence is counted to the total time period. If the proportion exceeds a preset value, it is initially judged as a downward trend. For a sequence initially identified as having an upward trend, its fluctuation amplitude is calculated. The fluctuation amplitude is the ratio of the difference between the maximum and minimum values ​​in the sequence to the mean of the sequence. If the fluctuation amplitude is less than the preset amplitude, the change trend type is determined to be continuous growth; if the fluctuation amplitude is greater than or equal to the preset amplitude, the change trend type is determined to be fluctuating growth. Calculate the fluctuation amplitude of the sequence initially judged to be a downward trend. If the fluctuation amplitude is less than a preset amplitude, the trend type is determined to be a continuous downward trend; if the fluctuation amplitude is greater than or equal to the preset amplitude, the trend type is determined to be a fluctuating downward trend.

7. The method according to claim 6, characterized in that, The process involves iterating through the membership values ​​of each level in the comprehensive evaluation vector, and determining the comprehensive evaluation result of the ecological restoration effect of steep slopes based on the level with the highest membership value. This includes: Extract the membership value corresponding to each level in the comprehensive evaluation vector. The membership value is used to indicate the degree to which the comprehensive evaluation result belongs to that level. Compare the membership values ​​of each level, determine the level with the largest membership value, and use this level as the preliminary evaluation result; Check the difference between the membership value corresponding to the preliminary evaluation result and the membership value of other levels. If the difference is less than the preset difference, then perform a second evaluation. When conducting a secondary evaluation, the weighted membership value of each level is calculated, and the weighted membership value is the product of the membership value and the reference score of that level. The weighted membership values ​​are summed and then divided by the total membership values ​​to obtain the comprehensive evaluation score. The corresponding evaluation level is determined based on the score range of the comprehensive evaluation score, and the comprehensive evaluation result of the ecological restoration effect of the steep slope is determined based on the evaluation level.

8. The method according to any one of claims 1-7, characterized in that, Based on the comprehensive evaluation results, a visual evaluation report is generated, including a time-axis evolution trend map, a spatial distribution heat map, and boundary lines for the classification of repair levels. The repair effect level data for different time periods are extracted from the comprehensive evaluation results, and the repair effect level data is associated with the corresponding timestamps to generate a timeline data sequence. A timeline evolution trend chart is plotted based on the timeline data sequence, and the level data at each time point is connected by a broken line; the horizontal axis of the timeline evolution trend chart is the timestamp, and the vertical axis is the repair effect level; Extract the geographic coordinates of each sampling point and the corresponding restoration effect level data from the slope restoration observation dataset to generate a spatial location-level dataset. Based on the spatial location-level dataset, an interpolation algorithm is used to generate spatially continuous repair effect level distribution data. The interpolation algorithm is used to convert discrete sampling point data into continuous spatial distribution data. A spatial distribution heatmap is drawn based on the spatially continuous repair effect level distribution data, and the color intensity of the spatial distribution heatmap is used to represent the high and low distribution of the repair effect level; The repair effect level data in the spatial distribution heat map is divided into regions to identify connected regions with the same repair effect level and extract the boundary contour lines of each connected region. The boundary contour lines are superimposed onto the spatial distribution heat map to generate repair level classification boundary lines; The timeline evolution trend chart, spatial distribution heat map, and repair level classification boundary line are integrated into the assessment report template to generate a visual assessment report that includes chart titles, coordinate axis descriptions, and legends.

9. The method according to any one of claims 1-7, characterized in that, The aforementioned method utilizes multi-source observation data from steep slope areas to form a slope restoration observation dataset with time-series correlation, including: Based on the interference of changes in illumination conditions on spectral reflectance, atmospheric correction and radiation normalization are performed on the surface cover spectral data to obtain the target spectral data. By extrapolating the trend of monitoring values ​​at adjacent time points, outlier identification and missing value imputation are performed on the soil environmental monitoring data to obtain continuous time-series soil environmental data. Based on the coordinates of ground control points, distortion correction and spatial registration processing are performed on the slope surface image data at different flight times to obtain spatially aligned surface image data. The vegetation physiological characteristic data are sample-verified to obtain the target physiological characteristic data; The target spectral data, the continuous temporal soil environmental data, the spatially aligned surface image data, and the target physiological characteristic data are correlated according to the acquisition timestamps to generate a slope restoration observation dataset containing data acquisition timestamps.

10. An evaluation system for ecological restoration of steep slopes, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of any one of the methods described in claims 1 to 9.

Images