A method for analyzing the correlation between root distribution characteristics and slope surface erosion

Abstract

The application provides a root system distribution feature and slope erosion correlation analysis method, comprising: acquiring root system image data, extracting root system distribution features from soil samples through a scanning device, separating root system structures by using an image segmentation algorithm, and obtaining a root system density distribution map; calculating a congestion degree index according to the root system density distribution map, grouping root system areas by using a clustering algorithm, and determining the numerical change of the congestion degree at different growth stages; for the stage division result, acquiring slope erosion simulation data, generating an erosion intensity distribution by using a water flow simulation software, and determining a corresponding matrix of the erosion intensity and the congestion degree; analyzing a non-linear corresponding relationship from the corresponding matrix, fitting a function relationship between the congestion degree and the erosion intensity by using a regression analysis model, and obtaining a fitting parameter set; inputting real-time root system data into an optimized corresponding relationship model to generate an erosion control effect prediction, and determining an optimal erosion prevention intervention point.

Description

Technical Field

[0001] This invention relates to the technical field of information technology, specifically to a method for analyzing the correlation between root distribution characteristics and slope erosion. Background Technology

[0002] Studying the relationship between root distribution characteristics and slope erosion is an important topic in ecological protection and land management. Slope erosion not only threatens the stability of soil resources but also has a profound impact on agricultural production and the ecological environment. Therefore, exploring the role of plant roots in controlling erosion is particularly crucial. By studying how roots influence soil structure and water flow paths, we can provide a scientific basis for erosion prevention measures, thereby ensuring the sustainable use of land.

[0003] However, current research methods have significant limitations in capturing the dynamic changes in the interaction between roots and slope erosion. Many studies tend to focus only on the root state at a fixed point in time, neglecting the continuous evolution of root characteristics during plant growth. This approach struggles to reveal the true impact of roots on soil conservation at different growth stages and fails to support precise timing and strategies for erosion control, resulting in limited effectiveness of related measures in practical applications.

[0004] A deeper technical challenge lies in the fact that the degree of root congestion is not static but continuously changes throughout the plant's growth cycle from germination to maturity, a characteristic that has not been fully considered. The degree of root congestion directly relates to the spatial distribution of roots in the soil and their ability to hold soil particles. Without accurately understanding its changing patterns, it is difficult to determine when the roots can truly stabilize the soil and mitigate erosion. Furthermore, this dynamic characteristic presents another challenge: the relationship between root congestion and slope erosion intensity is not a simple linear one, but rather influenced by a combination of factors such as growth time, increase in root quantity, and spatial expansion. For example, in the early stages of plant growth, the root system may be sparse and shallow, unable to effectively resist water erosion. Only after the number and depth of roots increase at a certain stage does soil stability significantly improve, but the exact point in time when this effect is achieved still lacks a clear standard for judgment.

[0005] Therefore, how to accurately track the changes in root congestion characteristics at different growth stages and clarify the correspondence between them and the effect of slope erosion control has become a key issue that needs to be addressed in this study. Summary of the Invention

[0006] This invention provides a method for analyzing the correlation between root distribution characteristics and slope erosion. The purpose is to solve the problem in the prior art of how to accurately track the changes in root congestion characteristics at different growth stages and clarify the correspondence between them and the slope erosion control effect.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A method for analyzing the correlation between root distribution characteristics and slope erosion, the method comprising:

[0009] Root image data is acquired, root distribution features are extracted from soil samples using a scanning device, and the root structure is separated using an image segmentation algorithm to obtain a root density distribution map.

[0010] Based on the root density distribution map, the congestion index was calculated, and the root regions were grouped using a clustering algorithm to determine the numerical changes in congestion at different growth stages.

[0011] Time series data is extracted from the numerical changes in congestion level, and a time series analysis model is used to predict the trend of congestion level. It is then determined whether the trend exceeds a preset threshold. If it does, it is marked as a high congestion stage, and the stage division results are obtained.

[0012] Based on the phase division results, slope erosion simulation data were obtained, and erosion intensity distribution was generated using water flow simulation software to determine the correspondence matrix between erosion intensity and blockage degree.

[0013] The nonlinear correspondences are analyzed from the correspondence matrix, and a regression analysis model is used to fit the functional relationship between the degree of congestion and the intensity of erosion, so as to obtain the set of fitting parameters.

[0014] Based on the set of fitted parameters, the stability of the functional relationship within the growth cycle is determined. If the stability is lower than the preset threshold, the parameters are adjusted and iterative optimization is performed to obtain the optimized corresponding relationship model.

[0015] By using the optimized correspondence model and inputting real-time root system data, we can generate a prediction of erosion control effectiveness and determine the optimal erosion prevention intervention point.

[0016] In one aspect of the invention, the acquisition of root image data, extracting root distribution features from soil samples using a scanning device, separating root structures using an image segmentation algorithm, and obtaining a root density distribution map includes:

[0017] Obtain root image data from soil samples;

[0018] The soil sample was imaged layer by layer using a scanning device to obtain the original root system image data;

[0019] An image segmentation algorithm is used to process the original root image data, separating the root pixel region from the soil background region to obtain a binary image of the root system.

[0020] The proportion of root system pixels to total pixels in each local region is calculated based on the binary root system image to obtain the local root system density value.

[0021] Spatial location mapping is performed on all local root density values ​​to generate a two-dimensional root density distribution map;

[0022] If there are areas in the root density distribution map where the density value is lower than the preset threshold, they are marked as hollow areas and their coordinates are recorded.

[0023] By comparing the coordinates of the cavity area with the surrounding density values, the missing root growth locations are determined and a list of missing locations is output.

[0024] In one aspect of the invention, the step of calculating a congestion level index based on a root density distribution map, grouping root regions using a clustering algorithm, and determining the numerical change of congestion level at different growth stages includes:

[0025] Obtain the root pixel percentage data for each local region in the root density distribution map;

[0026] The density distribution map is divided into multiple independent sub-regions by pre-defined grid division, and the congestion level index sequence of each sub-region is obtained.

[0027] Based on the congestion index sequence, the k-means clustering algorithm is used to group the regions, resulting in multiple cluster sets with different root congestion levels.

[0028] Extract the time stamp of the corresponding growth stage from each cluster set to obtain the time series congestion value of each cluster;

[0029] By comparing the congestion values ​​of adjacent growth stages, it can be determined whether the current cluster is experiencing an increase in congestion.

[0030] If the increase in congestion value in an adjacent phase exceeds the average increase in the value of adjacent clusters, then the cluster is marked as an intensified congestion cluster.

[0031] Extract the spatial coordinate range of the congestion-intensifying clusters in the density distribution map to determine the location of the congestion intensification.

[0032] Based on the comparison between the congestion value of the congestion aggravation location and the congestion value of the surrounding clusters, if the value of the surrounding clusters is consistently lower than that of the location, then the location is marked as a root congestion continuation region.

[0033] Output a list of coordinates for all regions where congestion is exacerbated and regions where congestion persists.

[0034] In one aspect of the invention, the step of extracting time-series data from changes in congestion levels, using a time-series analysis model to predict congestion trends, determining whether the trend exceeds a preset threshold, and marking it as a high-congestion stage if it does, to obtain stage division results, includes:

[0035] Numerical change data is obtained from the historical records of congestion levels and organized into a time series format to obtain a structured time series dataset.

[0036] For structured time series datasets, a time series analysis model is used to predict the trend direction, and the prediction results of the trend direction are obtained.

[0037] Based on the predicted trend direction, compare it with a preset threshold. If the predicted trend exceeds the preset threshold, it is marked as a high congestion state, and the marked state information is obtained.

[0038] By using the marked state information, different congestion stages are divided, the start and end times of each stage are determined, and a detailed list of stage divisions is obtained.

[0039] For the detailed list of phase divisions, extract the time periods corresponding to the high congestion state, determine whether there are consecutive high congestion time periods, and if there are consecutive time periods, mark them as key attention intervals to obtain the range of key attention intervals.

[0040] Based on the range of the key focus area, obtain the details of the changes in the congestion level within the corresponding time period, determine the fluctuations in the changes, and obtain descriptive data of the fluctuations.

[0041] By analyzing the descriptive data of fluctuations, the stability of numerical changes during high congestion phases is analyzed. If the fluctuations exceed the preset range, they are marked as unstable phases, and the final stability classification results are obtained.

[0042] In one aspect of the invention, the step of obtaining slope erosion simulation data based on the stage division results, generating erosion intensity distribution through water flow simulation software, and determining the correspondence matrix between erosion intensity and congestion degree includes:

[0043] Obtain historical data from slope erosion simulation and extract numerical sequences of erosion intensity.

[0044] For the numerical sequence of erosion intensity, an autoregressive integral moving average model is used for fitting to obtain the predicted trend sequence;

[0045] Based on the predicted trend sequence, determine whether each point in the sequence exceeds the preset intensity threshold. If it does, mark the high erosion stage to which the corresponding time point belongs, and obtain the start and end time marking results of the high erosion stage.

[0046] For the start and end time annotation results of the high erosion stage, the time period data corresponding to each high erosion stage is extracted from the slope erosion simulation data;

[0047] For the time period data corresponding to each high erosion stage, a spatial distribution map of erosion intensity is generated using water flow simulation software to obtain the spatial distribution results of erosion intensity for each stage;

[0048] Based on the spatial distribution results of erosion intensity at each stage and the numerical sequence of congestion degree within the corresponding time period, a correspondence matrix between erosion intensity and congestion degree is constructed to obtain a correspondence matrix shared by multiple stages.

[0049] For the corresponding matrix shared across multiple stages, it is determined whether the congestion level value increases synchronously when the erosion intensity value in the matrix increases. If it increases synchronously, it is identified as a positively correlated region, and a matrix subset of the positively correlated region is obtained.

[0050] In one aspect of the invention, the step of analyzing nonlinear correspondences from the correspondence matrix and fitting a regression analysis model to obtain a set of fitting parameters includes:

[0051] For the data in the corresponding matrix, the correlation data between congestion degree and erosion intensity are obtained, and a regression analysis model is used to process it to obtain a preliminary description of the functional relationship;

[0052] Based on the preliminary description of the functional relationship, the key fitting parameters are extracted, and the core numerical range of the parameter set is determined.

[0053] By constructing a prediction mapping table between congestion level and erosion intensity using the core numerical range of the parameter set, the corresponding entries in the prediction mapping table are obtained.

[0054] For the corresponding entries in the prediction mapping table, analyze the changing trends between entries, and if the changing trend exceeds the preset threshold range, mark it as an abnormal association interval, and obtain the distribution results of the abnormal association interval;

[0055] From the distribution results of abnormal correlation intervals, extract the high-frequency interval segment data and determine the distribution characteristics of the high-frequency interval segments;

[0056] Based on the distribution characteristics of high-frequency intervals, targeted data filtering rules are generated. If the filtering rules meet the preset conditions, the original data in the corresponding matrix is ​​processed into layers to obtain a layered data set.

[0057] By analyzing the differences between the data in each layer of the stratified dataset, the classification results of these differences are determined.

[0058] In one aspect of the present invention, the step of determining the stability of the functional relationship within the growth cycle based on the set of fitted parameters, and adjusting the parameters through iterative optimization if the stability is lower than a preset threshold, to obtain an optimized corresponding relationship model, includes:

[0059] Based on the fitting parameters and the functional relationship, the core data in the parameter set is obtained, and the parameter combination with a high degree of matching is determined by comparing the matching degree between the parameters and the functional relationship.

[0060] Based on the parameter combinations with high matching degree, we analyze their performance during the growth cycle and obtain a detailed description of the cycle performance by recording the data changes at each stage of the cycle.

[0061] By describing the periodic performance in detail, the stability value is compared with the preset threshold. If the stability value is lower than the preset threshold, the iterative adjustment process is triggered to determine the range of parameters after adjustment.

[0062] For the adjusted parameter range, parameter optimization is performed, and regression analysis model is used to correct the parameters round by round to obtain the optimized parameter set;

[0063] Based on the optimized parameter set, an updated version of the corresponding model is constructed. By analyzing the output results of the model under different periodic performance, the adaptability of the model improvement is judged.

[0064] Based on the adaptive results of the model improvement, and combined with the output of the relationship analysis, a targeted adjustment strategy is generated. This strategy is then applied to subsequent periodic data processing to determine the final model version.

[0065] Based on the final model version, its stability performance during the growth cycle is recorded. Through continuous data collection and comparison, a description of the functional relationship under long-term operation is obtained.

[0066] In one aspect of the present invention, the step of generating an erosion control effect prediction by inputting real-time root system data through an optimized correspondence model and determining the optimal anti-corrosion intervention point includes:

[0067] Obtain the real-time collected root morphology parameter sequence;

[0068] By processing the root morphology parameter sequence using a correspondence model, a sequence of predicted values ​​for erosion control effectiveness is obtained.

[0069] For the predicted value sequence of erosion control effect, we divide it into multiple consecutive time periods, calculate the statistical characteristics of the predicted values ​​in each time period, and obtain the set of time period distribution characteristics.

[0070] Based on the set of time period distribution characteristics, identify the intervals in which the predicted values ​​deviate from the normal range and determine a list of potential risk intervals;

[0071] From the list of potential risk intervals, extract the start and end times of each interval and the magnitude of change of the predicted value to determine the location of the risk interval;

[0072] If the risk interval is located in the critical growth period, then the interval is marked as the priority intervention interval, and the center time of the priority intervention interval is obtained;

[0073] Based on the center time of the priority intervention zone and the root morphology parameter sequence, determine the coordinates of the optimal anti-corrosion intervention point;

[0074] Based on the coordinates of the optimal anti-corrosion intervention point, and considering the current time and growth cycle stage, the time from the current moment to the intervention point coordinates is calculated to determine the timing of implementation.

[0075] Compared with the prior art, the present invention has the following beneficial effects:

[0076] This invention acquires root images of soil samples using a scanning device, extracts root density distribution maps using image segmentation algorithms, calculates congestion indices, and employs clustering algorithms to group root regions to quantify congestion trends at different growth stages. Subsequently, a time-series analysis model is used to predict the congestion time series, determining whether a high-congestion stage has been entered, thus achieving precise stage division within the growth cycle. Based on this, erosion intensity distribution data generated by slope flow simulation software is combined to construct a correspondence matrix between congestion degree and erosion intensity. A nonlinear function relationship is fitted through regression analysis, and the stability of the function throughout the entire growth cycle is evaluated. Unstable parameters are iteratively optimized, ultimately forming a high-precision optimized correspondence model. This model can directly predict erosion control effects and determine the optimal timing for erosion prevention intervention by inputting real-time root data. This invention deeply couples the dynamic evolution of root congestion with the slope erosion process, achieving a complete closed loop from root structure monitoring to erosion trend prediction and intervention optimization. It provides a scientific, precise, and intelligent prevention and control technology path for slope soil and water conservation, effectively improving the timeliness and targeting of erosion prevention and control. Attached Figure Description

[0077] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0078] Fig. 1 This is a flowchart of a method for analyzing the correlation between root distribution characteristics and slope erosion according to the present invention.

[0079] Fig. 2 This is a schematic diagram of a method for analyzing the correlation between root distribution characteristics and slope erosion according to the present invention.

[0080] Fig. 3 This is another schematic diagram of a method for analyzing the correlation between root distribution characteristics and slope erosion according to the present invention. Detailed Implementation

[0081] The present invention will be further described below with reference to embodiments. These embodiments are merely some, not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the protection scope of the present invention.

[0082] Please see Figs. 1-3 As shown, this embodiment discloses as follows: Figs. 1-3 This embodiment of a method for analyzing the correlation between root distribution characteristics and slope erosion may specifically include:

[0083] S101. Obtain root image data, extract root distribution features from soil samples using a scanning device, separate root structure using an image segmentation algorithm, and obtain root density distribution map.

[0084] Obtain root image data from soil samples. The soil sample is imaged layer by layer using a scanning device to obtain raw root image data. An image segmentation algorithm is used to process the raw root image data, separating the root pixel region from the soil background region to obtain a binary root image. The proportion of root pixels to total pixels in each local region is calculated based on the binary root image to obtain the local root density value. Spatial location mapping is performed on all local root density values ​​to generate a two-dimensional root density distribution map. If there are regions in the root density distribution map with density values ​​below a preset threshold, they are marked as void regions and their coordinates are recorded. The coordinates of the void regions are compared with the surrounding density values ​​to determine the locations of missing root growth, and a list of missing locations is output.

[0085] First, 3D imaging of collected soil columnar samples was performed using a CT scanner. The scanning voltage was set to 120kV, the current to 200mA, the slice thickness to 0.1mm, and the resolution to 50 micrometers, obtaining high-precision grayscale volumetric image data containing the root system and soil matrix. Then, a deep learning segmentation model based on an improved U-Net was used for root structure separation. This model uses ResNet-34 as the encoder backbone and incorporates an attention mechanism module to enhance the extraction of fine root features. During training, the Dice loss function combined with cross-entropy loss was used, across a total of 1200 labeled root slices. The data was trained on the dataset for 80 epochs, achieving an IoU of 0.87 on the validation set. After segmentation, a root binary mask was obtained. Noise points smaller than 0.05 mm³ were removed and the true root structure was preserved through stereoscopic 3D connected component analysis. Then, the entire soil sample was divided into stereoscopic grids using 0.5 mm × 0.5 mm × 0.5 mm as statistical units. The proportion of root voxels to the total volume of each cell was calculated as the local root density value. Finally, a 3D Gaussian filter (sigma = 1.2) was used to smooth the density field to reduce the influence of discrete noise. A 3D visualization of root density distribution was generated using the marching cubes algorithm with a density threshold of 0.003 mm³ / mm³ as the isosurface. Simultaneously, the average root density was calculated every 5 mm along the soil depth direction, and a depth-density curve was plotted to analyze the root distribution patterns in the 0-30 cm tillage layer, 30-60 cm subsurface layer, and below 60 cm. This completed the entire process of quantitative extraction and characterization of root density distribution characteristics from raw scan data.

[0086] S102. Based on the root density distribution map, calculate the congestion index, use a clustering algorithm to group the root regions, and determine the numerical changes in congestion at different growth stages.

[0087] Obtain the root pixel percentage data for each local region in the root density distribution map. Divide the density distribution map into multiple independent sub-regions using a pre-defined grid, obtaining a congestion level index sequence for each sub-region. Use k-means clustering to group regions based on the congestion level index sequence, resulting in multiple cluster sets with different root congestion levels. Extract the time stamp corresponding to the growth stage from each cluster set to obtain the time-series congestion value for each cluster. By comparing the congestion values ​​of adjacent growth stages, determine if the current cluster is experiencing an increase in congestion. If the increase in congestion value in an adjacent stage exceeds the average increase of adjacent clusters, mark the cluster as an aggravated congestion cluster. Extract the spatial coordinate range of the aggravated congestion cluster in the density distribution map to determine the location of the aggravated congestion. Based on the comparison of the congestion values ​​of the aggravated congestion location with those of surrounding clusters, if the values ​​of surrounding clusters are consistently lower than that of this region, mark this location as a persistent root congestion region. Output a list of coordinates for all aggravated congestion location regions and persistent congestion regions.

[0088] Based on the obtained root density distribution map, the root congestion index within each stereoscopic grid cell was first calculated. This index was defined as the ratio of the number of root voxels within the cell to the maximum number of root voxels the cell could accommodate. The maximum capacity was calculated based on a soil porosity of 0.42 and an average cross-sectional area of ​​0.008 mm², resulting in a congestion index ranging from 0 to 1. Subsequently, the congestion indices of all cells were used as feature vectors, and K-means clustering was employed to group the regions. The number of clusters was set to 4, initial centers were selected using the k-means++ method, the number of iterations was limited to 30, and Euclidean distance was used as the distance metric. The root regions were ultimately divided into four categories: extremely low congestion (index 0-0.15), low congestion (0.15-0.35), moderate congestion (0.35-0.60), and high congestion (above 0.60). Then, the above clustering process was performed on the independent density distribution data for each growth stage, for example, at the four key stages of seedling, jointing, heading, and maturity. Samples were collected for each stage, and the percentage of volume occupied by each type of cluster was calculated, along with the average congestion index for each type of cluster. During the analysis, the volume percentage changes of highly congested areas in the 0-20cm shallow layer, 20-50cm middle layer, and below 50cm deep layer were statistically analyzed. It was found that the proportion of highly congested areas in the seedling stage was only 2.8%, mainly concentrated in the 0-10cm topsoil layer, while this proportion rose to 18.6% by the heading stage and extended downwards to a depth of 40cm, indicating that the root system gradually forms a densely intertwined structure locally as it grows and develops. Finally, the volume percentage of each type of cluster and the average congestion index at each stage were combined into a time-series matrix. By calculating the percentage increase in volume of highly congested areas and the rate of change of the mean index between adjacent stages, the evolution trend of congestion degree was quantified. For example, the increase in highly congested volume from the jointing stage to the heading stage reached 124%, and the average index rose from 0.58 to 0.71, thus revealing the dynamic law of significantly aggravated root congestion during the reproductive growth stage of crops, providing a quantitative basis for subsequent water and nutrient transport simulation and stress resistance evaluation.

[0089] S103. Extract time series data from the numerical changes in congestion level, use a time series analysis model to predict the trend of congestion level, determine whether the trend exceeds a preset threshold, and if it does, mark it as a high congestion stage to obtain the stage division results.

[0090] Numerical change data is extracted from historical records of congestion levels and organized into a time series format to obtain a structured time series dataset. For this structured dataset, a time series analysis model is used to predict the trend direction. Based on the predicted trend direction, a preset threshold is compared. If the predicted trend exceeds the preset threshold, it is marked as a high-congestion state, yielding the marked state information. Using this marked state information, different congestion stages are identified, and the start and end times of each stage are determined, resulting in a detailed list of stage divisions. From this detailed list, the time periods corresponding to the high-congestion states are extracted, and it is determined whether consecutive high-congestion time periods exist. If consecutive time periods exist, they are marked as key focus intervals, yielding the range of these key focus intervals. Based on the range of these key focus intervals, details of the numerical changes in congestion levels within the corresponding time periods are obtained, determining the fluctuation patterns and obtaining descriptive data on these fluctuations. Using this descriptive data, the stability of numerical changes within the high-congestion stages is analyzed. If the fluctuations exceed a preset range, it is marked as an unstable stage, yielding the final stability classification result.

[0091] Time series data were extracted from the volume percentage and average congestion index sequences of highly congested areas at each growth stage. Using the seedling stage, jointing stage, heading stage, and maturity stage as nodes, highly congested volume percentage sequences of 2.8%, 9.1%, 18.6%, and 21.4% and average congestion index sequences of 0.62, 0.58, 0.71, and 0.69 were constructed. The volume percentage sequences were fitted using the ARIMA time series analysis model. First, the ADF test confirmed that the sequence was stationary after first differencing, and the model parameters were determined to be ARIMA(1,1,1). The model coefficients were obtained using maximum likelihood estimation and subjected to residual white noise testing. Subsequently, rolling predictions were made for the two potential late-stage growth phases, yielding predicted volume percentages of 23.8% and 26.1%. The prediction results were then analyzed. Compared with the preset high congestion warning threshold of 20%, it was found that the heading stage had exceeded the threshold, and the predicted value in subsequent stages further increased. Therefore, the heading stage and thereafter were marked as the high congestion stage. To enhance the reliability of trend judgment, the exponential smoothing method (Holt linear trend model, smoothing parameters α=0.3, β=0.1) was applied to the average congestion index sequence. The smoothed trend line showed that the index continued to rise from the jointing stage and was predicted to stabilize at around 0.73 in the later stage, confirming the persistence of the high congestion stage. Finally, the time points when the volume ratio exceeded the threshold and the direction of the index trend were integrated, and the stage division results were output as follows: the seedling stage and jointing stage belong to the normal congestion stage, the heading stage enters the high congestion stage and continues to the maturity stage. This division result can be directly input into the root water and nutrient competition simulation module for dynamic adjustment of deep nutrient supply strategy.

[0092] S104. Based on the stage division results, obtain slope erosion simulation data, generate erosion intensity distribution through water flow simulation software, and determine the correspondence matrix between erosion intensity and congestion degree.

[0093] Historical data from slope erosion simulations were acquired, and erosion intensity numerical sequences were extracted. An autoregressive integral moving average model was used to fit the erosion intensity numerical sequences to obtain a predicted trend sequence. Based on the predicted trend sequence, it was determined whether each point in the sequence exceeded a pre-set intensity threshold. If it did, the corresponding time point was marked as belonging to a high erosion stage, obtaining the start and end time labeling results for the high erosion stage. For the start and end time labeling results of the high erosion stages, time period data corresponding to each high erosion stage were extracted from the slope erosion simulation data. For the time period data corresponding to each high erosion stage, a spatial distribution map of erosion intensity was generated using water flow simulation software, obtaining the spatial distribution results of erosion intensity for each stage. A correspondence matrix between erosion intensity and congestion degree was constructed based on the spatial distribution results of erosion intensity for each stage and the corresponding congestion degree numerical sequence within the time period, obtaining a multi-stage shared correspondence matrix. For the multi-stage shared correspondence matrix, it was determined whether the congestion degree value increased synchronously with the increase in erosion intensity value. If they increased synchronously, it was identified as a positively correlated region, obtaining a matrix subset of positively correlated regions.

[0094] First, based on topographic elevation data and rainfall sequences, the RUSLE model combined with the surface runoff accumulation algorithm was used to divide the slope erosion stages, classifying the slope into the initial sheet erosion stage (erosion modulus less than 2000 t / km²·a), the gully erosion development stage (2000~8000 t / km²·a), and the severe gully erosion stage (greater than 8000 t / km²·a). Then, using the HEC-RAS two-dimensional hydrodynamic model, the rainfall process line of a 60-minute design storm with a 30 mm / h flow rate and slope parameters with a Manning roughness of 0.035 were input to simulate and generate water depth, flow velocity, and shear stress distribution maps for each grid cell. Subsequently, erosion intensity raster data with a spatial resolution of 5 m was calculated using a modified general soil loss equation, with the maximum erosion intensity reaching 12.7 kg / m² and the average value being 3.4 kg / m². The erosion intensity values ​​were then further divided into three stages: 0~1 kg / m², 1~3 kg / m², 3~7 kg / m², and 7~15 kg / m². The slope congestion was classified into four levels based on the simulated sediment deposition thickness: no congestion (0-2 cm), slight congestion (2-8 cm), moderate congestion (8-20 cm), and severe congestion (greater than 20 cm). A 4×4 transition probability matrix was constructed by statistically analyzing the correspondence of more than 50,000 grid cells. For example, the probability of no congestion was 0.82 for the erosion intensity range of 0-1 kg / m², 0.15 for slight congestion, and 0.03 for moderate and above congestion. In contrast, the probability of severe congestion for the 7-15 kg / m² range increased to 0.67, and the combined probability of slight and moderate congestion was 0.31. This formed a quantitative correspondence matrix between erosion intensity and congestion degree, providing data support for subsequent slope management priority ranking and dynamic risk assessment. The entire process used Python to call GDAL and NumPy to achieve batch reclassification of raster data and matrix statistics, ensuring logical consistency and repeatability of results.

[0095] S105. Analyze the nonlinear correspondence from the correspondence matrix, and use a regression analysis model to fit the functional relationship between the degree of congestion and the intensity of erosion to obtain the set of fitting parameters.

[0096] For the data in the corresponding matrix, the correlation data between congestion level and erosion intensity is obtained. A regression analysis model is used to process this data, resulting in a preliminary functional relationship description. Based on this preliminary functional relationship description, key fitting parameters are extracted, and the core numerical range of the parameter set is determined. Using the core numerical range of the parameter set, a prediction mapping table of congestion level and erosion intensity is constructed, and the corresponding entries in the prediction mapping table are obtained. For the corresponding entries in the prediction mapping table, the changing trends between entries are analyzed. If the changing trend exceeds a preset threshold range, it is marked as an abnormal correlation interval, and the distribution results of the abnormal correlation intervals are obtained. From the distribution results of the abnormal correlation intervals, high-frequency interval data is extracted, and the distribution characteristics of high-frequency intervals are determined. Based on the distribution characteristics of high-frequency intervals, targeted data filtering rules are generated. If the filtering rules meet preset conditions, the original data in the corresponding matrix is ​​stratified, resulting in a stratified dataset. Using the stratified dataset, the differences between the data in each layer are analyzed, and the classification results of the differences are determined.

[0097] Based on the aforementioned 4×4 transition probability matrix of erosion intensity and congestion degree, nonlinear correspondence features were further extracted. A combination of polynomial regression and quantile regression was used to fit the functional relationship between congestion thickness and erosion intensity. First, erosion intensity was used as the independent variable x (unit: kg / m²), and congestion thickness as the dependent variable y (unit: cm). After cleaning over 50,000 grid samples and removing outliers (points where y > 80 cm or x > 20 kg / m²), approximately 48,000 valid samples were retained. Subsequently, a cubic polynomial regression model y = a·x³ + b·x² + c·x + d was constructed, supplemented by quantile regression (τ = 0.5, 0.75, 0.90) to capture the conditional distribution under different congestion risk levels. The fitting process was implemented using the statsmodels library in Python. The results showed that the cubic polynomial achieved an R² of 0.714 in the overall fit, with parameters a = 0.042, b = 0.319, c = 2.847, and d = 0.956, indicating that for every 1 increase in erosion intensity... At erosion intensity of 4.5 kg / m², the congestion thickness shows an accelerated growth trend, especially when the erosion intensity exceeds 4.5 kg / m², the slope of the curve increases significantly; quantile regression further reveals that when the erosion intensity is 5 kg / m², the median congestion thickness is approximately 4.2 cm, the 75th quantile reaches 9.8 cm, and the 90th quantile jumps to 18.6 cm, reflecting the dispersion of congestion intensity in high-erosion areas and the amplification effect of extreme risks; to improve the practicality of the model in governance decision-making, spatial autocorrelation analysis (Moran's) was performed on the fitting residuals. With I=0.27 and p<0.001, significant spatial clustering was confirmed. Therefore, geographically weighted regression (GWR) was introduced as a supplement. Adaptive kernel function and AICc criterion were used to optimize bandwidth, and finally, a set of local regression parameters was obtained. The regression coefficient α value in the high-altitude steep slope area can reach 0.058, while the α value in the low-slope gentle slope area drops to 0.031, showing significant differences in spatial heterogeneity parameters. These fitted parameter sets and quantile curves together constitute a quantitative functional relationship, providing continuous and differentiable mathematical support for subsequent prediction of congestion thickness based on erosion intensity, estimation of remediation engineering volume, and dynamic threshold early warning. The entire analysis process was automated by using Python's statsmodels, mgwr, and PySAL libraries, ensuring the traceability and consistency of the results.

[0098] S106. Based on the set of fitted parameters, determine the stability of the functional relationship within the growth cycle. If the stability is lower than the preset threshold, adjust the parameters and iteratively optimize to obtain the optimized corresponding relationship model.

[0099] For the fitted parameters and functional relationships, core data from the parameter set is obtained. By comparing the matching degree between the parameters and the functional relationship, parameter combinations with high matching degree are determined. Based on the parameter combinations with high matching degree, their performance within the growth cycle is analyzed. Data changes at each stage of the cycle are recorded to obtain a detailed description of the cycle performance. Through this detailed description of cycle performance, stability values ​​are extracted and compared with preset thresholds. If the stability value is lower than the preset threshold, an iterative adjustment process is triggered to determine the adjusted parameter range. For the adjusted parameter range, parameter optimization is implemented. A regression analysis model is used to correct the parameters round by round to obtain an optimized parameter set. Based on the optimized parameter set, an updated version of the corresponding model is constructed. By analyzing the model's output under different cycle performances, the adaptability of the model improvement is judged. Based on the adaptability results of the model improvement, combined with the output of the relationship analysis, a targeted adjustment strategy is generated. This strategy is applied to subsequent cycle data processing to determine the final model version. Based on the final model version, its stability performance within the growth cycle is recorded. Through continuous data collection and comparison, a description of the functional relationship under long-term operation is obtained.

[0100] For the stability analysis and optimization of the functional relationship between erosion intensity and congestion degree, the stability index of the functional relationship within the growth cycle is first calculated using the obtained set of fitted parameters. Specifically, time series data is used, dividing the cycle into 10 time periods. For each period, approximately 5000 sample points of erosion intensity x (unit kg / m²) and congestion degree y (unit m) are collected. The volatility of the local regression parameters in each time period is calculated. Assuming the initial parameters are a=0.035, b=0.28, c=2.5, and d=1.1, the mean square error of the parameters in each time period is 0.012 using a self-developed stability evaluation algorithm in Python. If the preset stability threshold is 0.01, the stability is deemed insufficient, and parameter adjustment is required. Next, gradient descent was used for iterative optimization with a learning rate of 0.001 and 500 iterations. The parameters were gradually adjusted by minimizing the mean squared error (MSE) between the predicted and observed values. The parameter changes were recorded at each iteration, resulting in the optimized parameter set: a = 0.038, b = 0.25, c = 2.6, and d = 1.05. The MSE decreased from the initial 0.85 to 0.62, indicating improved model prediction accuracy. To ensure the optimized model adapts to different terrain conditions, terrain slope was introduced as an auxiliary variable. An automated script calculated the weight of slope on the parameters. In areas with a slope greater than 15 degrees, parameter a was increased to 0.041, while in areas with a slope less than 5 degrees, a was decreased to 0.034, forming a dynamic parameter adjustment mechanism. Finally, the stability indexes for each time period within the growth cycle were recalculated using the optimized model. The mean squared error was reduced to 0.008, which is below the threshold of 0.01, confirming that the model's stability met the standard. The entire process was automated using Python scripts and matrix operations were performed using the NumPy library to ensure computational efficiency and consistency of results.

[0101] S107. Using the optimized correspondence model, real-time root data is input to generate a prediction of erosion control effect and determine the optimal anti-corrosion intervention point.

[0102] Acquire real-time root morphology parameter sequences. Process these sequences using a correspondence model to obtain a sequence of predicted erosion control effects. Divide the predicted erosion control effect sequence into multiple consecutive time periods, calculate the statistical characteristics of the predicted values ​​within each time period, and obtain a set of time period distribution characteristics. Based on this set, identify intervals where predicted values ​​deviate from the normal range, and determine a list of potential risk intervals. From this list, extract the start and end times and the magnitude of predicted value change for each interval to determine its location. If a risk interval is located during a critical growth period, mark it as a priority intervention interval and obtain its center time. Based on the center time of the priority intervention interval and the root morphology parameter sequence, determine the coordinates of the optimal erosion control intervention point. For the optimal intervention point coordinates, combine the current time and growth cycle stage to calculate the time from the current time to the intervention point coordinates, thus determining the implementation timing.

[0103] Using an optimized model relating erosion intensity to congestion level, and incorporating real-time root growth monitoring data, the effectiveness of erosion control was predicted. First, an IoT sensor network automatically acquired current root biomass density data (g / m³) and root distribution depth data (cm) every 30 minutes. Using 480 sets of real-time root data collected in the last 24 hours as input vectors, these were substituted into the optimized model y=f(x; a=0.038, b=0.25, c=2.6, d=1.05) to calculate the predicted erosion intensity and congestion level values ​​at the corresponding time points. Then, a forward time window sliding prediction method was employed, projecting 12 time points over the next 72 hours, with each point representing a 6-hour prediction. An incremental root biomass simulation sequence (generated based on a historical growth rate exponential model, with an average growth rate of 0.018 g / m³·h) was input point by point, yielding the erosion intensity sequence (range 0.12~0.47 kg / m²) and congestion level sequence (range 0.12~0.47 kg / m²) for each prediction node. The predicted sequence is then cross-referenced and analyzed by setting an erosion control threshold of 0.35 kg / m² and a congestion warning threshold of 3.0 m. When a certain time node simultaneously meets the conditions of erosion intensity greater than 0.35 kg / m² and congestion degree greater than 3.0 m, it is marked as a high-risk intervention point. The comprehensive risk score of each high-risk point is further calculated using the weighted formula: score = 0.6 × (erosion intensity / 0.35) + 0.4 × (congestion degree / 3.0). The highest-scoring node is the 7th predicted point (corresponding to 42 hours later), with a score of 1.78. The system automatically determines this node as the optimal erosion prevention intervention point and generates an intervention suggestion time window (within 36 to 48 hours in the future). At the same time, it outputs the corresponding root biomass density critical value of approximately 12.6 g / m³ as the trigger condition. The entire prediction and decision-making process is completed in real time in the background through a preset automated inference engine combined with Pandas data processing and SciPy numerical calculation modules, ensuring the accuracy and operability of the intervention timing.

[0104] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, the personal information processing rules are clearly informed through signs / information, and authorization is obtained through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing the correlation between root distribution characteristics and slope erosion, characterized in that, The method includes: Root image data is acquired, root distribution features are extracted from soil samples using a scanning device, and the root structure is separated using an image segmentation algorithm to obtain a root density distribution map. Based on the root density distribution map, the congestion index was calculated, and the root regions were grouped using a clustering algorithm to determine the numerical changes in congestion at different growth stages. Time series data is extracted from the numerical changes in congestion level, and a time series analysis model is used to predict the trend of congestion level. It is then determined whether the trend exceeds a preset threshold. If it does, it is marked as a high congestion stage, and the stage division results are obtained. Based on the phase division results, slope erosion simulation data were obtained, and erosion intensity distribution was generated using water flow simulation software to determine the correspondence matrix between erosion intensity and blockage degree. The nonlinear correspondences are analyzed from the correspondence matrix, and a regression analysis model is used to fit the functional relationship between the degree of congestion and the intensity of erosion, so as to obtain the set of fitting parameters. Based on the set of fitted parameters, the stability of the functional relationship within the growth cycle is determined. If the stability is lower than the preset threshold, the parameters are adjusted and iterative optimization is performed to obtain the optimized corresponding relationship model. By using the optimized correspondence model and inputting real-time root system data, we can generate a prediction of erosion control effectiveness and determine the optimal erosion prevention intervention point.

2. The method for analyzing the correlation between root distribution characteristics and slope erosion according to claim 1, characterized in that, The process of acquiring root image data involves extracting root distribution features from soil samples using a scanning device, separating the root structure using an image segmentation algorithm, and obtaining a root density distribution map, including: Obtain root image data from soil samples; The soil sample was imaged layer by layer using a scanning device to obtain the original root system image data; An image segmentation algorithm is used to process the original root image data, separating the root pixel region from the soil background region to obtain a binary image of the root system. The proportion of root system pixels to total pixels in each local region is calculated based on the binary root system image to obtain the local root system density value. Spatial location mapping is performed on all local root density values ​​to generate a two-dimensional root density distribution map; If there are areas in the root density distribution map where the density value is lower than the preset threshold, they are marked as hollow areas and their coordinates are recorded. By comparing the coordinates of the cavity area with the surrounding density values, the missing root growth locations are determined and a list of missing locations is output.

3. The method for analyzing the correlation between root distribution characteristics and slope erosion according to claim 1, characterized in that, The process of calculating congestion index based on root density distribution map, grouping root regions using clustering algorithm, and determining the numerical change of congestion level at different growth stages includes: Obtain the root pixel percentage data for each local region in the root density distribution map; The density distribution map is divided into multiple independent sub-regions by pre-defined grid division, and the congestion level index sequence of each sub-region is obtained. Based on the congestion index sequence, the k-means clustering algorithm is used to group the regions, resulting in multiple cluster sets with different root congestion levels. Extract the time stamp of the corresponding growth stage from each cluster set to obtain the time series congestion value of each cluster; By comparing the congestion values ​​of adjacent growth stages, it can be determined whether the current cluster is experiencing an increase in congestion. If the increase in congestion value in an adjacent phase exceeds the average increase in the value of adjacent clusters, then the cluster is marked as an intensified congestion cluster. Extract the spatial coordinate range of the congestion-intensifying clusters in the density distribution map to determine the location of the congestion intensification. Based on the comparison between the congestion value of the congestion aggravation location and the congestion value of the surrounding clusters, if the value of the surrounding clusters is consistently lower than that of the location, then the location is marked as a root congestion continuation region. Output a list of coordinates for all regions where congestion is exacerbated and regions where congestion persists.

4. The method for analyzing the correlation between root distribution characteristics and slope erosion according to claim 1, characterized in that, The process involves extracting time-series data from changes in congestion levels, using a time-series analysis model to predict congestion trends, determining whether the trend exceeds a preset threshold, and marking it as a high-congestion stage if it does, thus obtaining stage division results, including: Numerical change data is obtained from the historical records of congestion levels and organized into a time series format to obtain a structured time series dataset. For structured time series datasets, a time series analysis model is used to predict the trend direction, and the prediction results of the trend direction are obtained. Based on the predicted trend direction, compare it with a preset threshold. If the predicted trend exceeds the preset threshold, it is marked as a high congestion state, and the marked state information is obtained. By using the marked state information, different congestion stages are divided, the start and end times of each stage are determined, and a detailed list of stage divisions is obtained. For the detailed list of phase divisions, extract the time periods corresponding to the high congestion state, determine whether there are consecutive high congestion time periods, and if there are consecutive time periods, mark them as key attention intervals to obtain the range of key attention intervals. Based on the range of the key focus area, obtain the details of the changes in the congestion level within the corresponding time period, determine the fluctuations in the changes, and obtain descriptive data of the fluctuations. By analyzing the descriptive data of fluctuations, the stability of numerical changes during high congestion phases is analyzed. If the fluctuations exceed the preset range, they are marked as unstable phases, and the final stability classification results are obtained.

5. The method for analyzing the correlation between root distribution characteristics and slope erosion according to claim 1, characterized in that, Based on the stage division results, slope erosion simulation data is obtained, and erosion intensity distribution is generated using water flow simulation software to determine the correspondence matrix between erosion intensity and blockage degree, including: Obtain historical data from slope erosion simulation and extract numerical sequences of erosion intensity. For the numerical sequence of erosion intensity, an autoregressive integral moving average model is used for fitting to obtain the predicted trend sequence; Based on the predicted trend sequence, determine whether each point in the sequence exceeds the preset intensity threshold. If it does, mark the high erosion stage to which the corresponding time point belongs, and obtain the start and end time marking results of the high erosion stage. For the start and end time annotation results of the high erosion stage, the time period data corresponding to each high erosion stage is extracted from the slope erosion simulation data; For the time period data corresponding to each high erosion stage, a spatial distribution map of erosion intensity is generated using water flow simulation software to obtain the spatial distribution results of erosion intensity for each stage; Based on the spatial distribution results of erosion intensity at each stage and the numerical sequence of congestion degree within the corresponding time period, a correspondence matrix between erosion intensity and congestion degree is constructed to obtain a correspondence matrix shared by multiple stages. For the corresponding matrix shared across multiple stages, it is determined whether the congestion level value increases synchronously when the erosion intensity value in the matrix increases. If it increases synchronously, it is identified as a positively correlated region, and a matrix subset of the positively correlated region is obtained.

6. The method for analyzing the correlation between root distribution characteristics and slope erosion according to claim 1, characterized in that, The process involves analyzing the nonlinear correspondence from the correspondence matrix and using a regression analysis model to fit the functional relationship between congestion degree and erosion intensity, resulting in a set of fitting parameters, including: For the data in the corresponding matrix, the correlation data between congestion degree and erosion intensity are obtained, and a regression analysis model is used to process it to obtain a preliminary description of the functional relationship; Based on the preliminary description of the functional relationship, the key fitting parameters are extracted, and the core numerical range of the parameter set is determined. By constructing a prediction mapping table between congestion level and erosion intensity using the core numerical range of the parameter set, the corresponding entries in the prediction mapping table are obtained. For the corresponding entries in the prediction mapping table, analyze the changing trends between entries, and if the changing trend exceeds the preset threshold range, mark it as an abnormal association interval, and obtain the distribution results of the abnormal association interval; From the distribution results of abnormal correlation intervals, extract the high-frequency interval segment data and determine the distribution characteristics of the high-frequency interval segments; Based on the distribution characteristics of high-frequency intervals, targeted data filtering rules are generated. If the filtering rules meet the preset conditions, the original data in the corresponding matrix is ​​processed into layers to obtain a layered data set. By analyzing the differences between the data in each layer of the stratified dataset, the classification results of these differences are determined.

7. The method for analyzing the correlation between root distribution characteristics and slope erosion according to claim 1, characterized in that, The step involves determining the stability of the functional relationship within the growth cycle based on the fitted parameter set. If the stability is lower than a preset threshold, the parameters are adjusted and iteratively optimized to obtain the optimized corresponding relationship model, including: Based on the fitting parameters and the functional relationship, the core data in the parameter set is obtained, and the parameter combination with a high degree of matching is determined by comparing the matching degree between the parameters and the functional relationship. Based on the parameter combinations with high matching degree, we analyze their performance during the growth cycle and obtain a detailed description of the cycle performance by recording the data changes at each stage of the cycle. By describing the periodic performance in detail, the stability value is compared with the preset threshold. If the stability value is lower than the preset threshold, the iterative adjustment process is triggered to determine the range of parameters after adjustment. For the adjusted parameter range, parameter optimization is performed, and regression analysis model is used to correct the parameters round by round to obtain the optimized parameter set; Based on the optimized parameter set, an updated version of the corresponding model is constructed. By analyzing the output results of the model under different periodic performance, the adaptability of the model improvement is judged. Based on the adaptive results of the model improvement, and combined with the output of the relationship analysis, a targeted adjustment strategy is generated. This strategy is then applied to subsequent periodic data processing to determine the final model version. Based on the final model version, its stability performance during the growth cycle is recorded. Through continuous data collection and comparison, a description of the functional relationship under long-term operation is obtained.

8. The method for analyzing the correlation between root distribution characteristics and slope erosion according to claim 1, characterized in that, The process involves using an optimized correspondence model, inputting real-time root system data to generate a prediction of erosion control effectiveness, and determining the optimal erosion prevention intervention point, including: Obtain the real-time collected root morphology parameter sequence; By processing the root morphology parameter sequence using a correspondence model, a sequence of predicted values ​​for erosion control effectiveness is obtained. For the predicted value sequence of erosion control effect, we divide it into multiple consecutive time periods, calculate the statistical characteristics of the predicted values ​​in each time period, and obtain the set of time period distribution characteristics. Based on the set of time period distribution characteristics, identify the intervals in which the predicted values ​​deviate from the normal range and determine a list of potential risk intervals; From the list of potential risk intervals, extract the start and end times of each interval and the magnitude of change of the predicted value to determine the location of the risk interval; If the risk interval is located in the critical growth period, then the interval is marked as the priority intervention interval, and the center time of the priority intervention interval is obtained; Based on the center time of the priority intervention zone and the root morphology parameter sequence, determine the coordinates of the optimal anti-corrosion intervention point; Based on the coordinates of the optimal anti-corrosion intervention point, and considering the current time and growth cycle stage, the time from the current moment to the intervention point coordinates is calculated to determine the timing of implementation.

Images