A method for roadside slope vegetation restoration growth and construction

Abstract

The application discloses a method for highway slope vegetation recovery and construction, and belongs to the field of slope greening and ecological environmental protection. Based on the investigation of the built highway slope vegetation coverage and the basic engineering condition indexes and the plant growth basic construction indexes, the indexes are sequentially optimized by using the field investigation and image recognition method, an optimization-prediction-reoptimization-reprediction mode is formed, the basic construction indexes of the slope vegetation recovery are continuously optimized until the slope vegetation recovery effect reaches the good or better grade, and the optimization scheme of the slope recovery process is proposed for indexes which cannot be optimized, such as the slope direction, the slope position, the slope position and the altitude. The method for the highway slope vegetation recovery and construction can optimize the highway slope greening mode, screen the applicable ecological recovery method and improve the slope ecological recovery effect.

Description

Technical Field

[0001] This application belongs to the field of slope greening and ecological environmental protection, and specifically relates to a method for the restoration and construction of vegetation on highway slopes. Background Technology

[0002] The construction of infrastructure such as highways and railways inevitably encroaches on and damages vegetation along the route, and restoration is extremely difficult once damaged. In the construction of highways in this area, turf transplantation technology has been widely applied. This involves removing, dividing, and storing the turf occupied by highways or railways, and then re-laying it during highway slope restoration. This approach minimizes the damage to the local ecological environment caused by construction and restores the roadside ecological environment at a relatively low cost through re-laying.

[0003] However, in engineering practice, the design of turf laying processes is not entirely reasonable, mainly reflected in the following aspects: First, the selection of turf laying locations is inappropriate, such as laying on steep excavated slopes or within arched frameworks or frame beams, leading to large areas of turf drying and dying in water-scarce and soil-poor environments. Second, the cultivation of turf and the underlying base soil is not carried out simultaneously, especially on some rocky or dry-laid stone slopes. Under traditional manual laying methods, the bond between turf and the underlying soil, and between the soil and the underlying rock, is not tight enough, preventing the turf root system from effectively fixing itself to the soil and absorbing water and nutrients, resulting in gradual drying and death after a period of time. Third, the presence of intercepting or drainage ditches above the turf-laying slope prevents runoff from replenishing the turf with water, causing turf patches to dry out. In addition, some turf is used in different altitudes and climate zones, making it difficult to adapt to the environment of the laying area and causing it to die.

[0004] In the existing technology, there is no decision-making technology for the construction of the foundation for vegetation restoration and growth on highway slopes. As a result, when implementing turf laying on highway slopes, it is impossible to estimate the effect of turf laying under different slope foundation engineering design indicators and plant growth foundation construction indicators in advance. Therefore, it is impossible to guide the design of slope vegetation restoration measures, and cannot assist in the design optimization of slope ecological restoration measures, the selection of restoration measures and application decisions.

[0005] Therefore, in order to predict and make decisions regarding the restoration of vegetation on highway slopes, it is urgent to find the influencing factors affecting the construction of vegetation restoration on slopes, and to establish a predictive model for the survival rate of turf laying on highway slopes. By repeatedly inputting optimized slope foundation engineering design indicators and plant growth foundation construction indicators into the model, the effect of improving the survival rate of turf laying on slopes can be simulated, providing decision-making assistance for the design of slope vegetation restoration projects and the selection of restoration measures, improving the effect of turf laying on slopes in future highway construction, and benefiting the vegetation restoration work on highway slopes. Summary of the Invention

[0006] The main purpose of this application is to overcome the shortcomings of existing technologies, such as the lack of decision-making technology for the construction of the foundation for vegetation restoration and growth on highway slopes, the inability to predict the effect of turf laying under different slope foundation engineering design indicators and plant growth foundation construction indicators, and the inability to assist in the selection and application of ecological restoration processes. The technical problem to be solved is to establish a general model for predicting the effect of turf laying on slopes, which can be widely applied to the prediction of the effect of turf laying on highway slopes.

[0007] Another objective of this application is to overcome the shortcomings of existing engineering practices, such as the inability of highway slope foundation engineering design conditions to adapt to slope turf laying, the lack of targeted selection of plant growth foundation construction measures, and the inability to guarantee the ecological effect of slope turf laying. The technical problem to be solved is to make the selection of highway slope turf laying technology more in line with the slope foundation engineering design conditions, supplemented by plant growth foundation construction technology measures, to improve the survival rate of slope turf, thereby providing support for the optimization of engineering design indicators and plant growth foundation construction decisions, and improving the ecological restoration effect of slope turf laying.

[0008] This application provides a method for vegetation restoration and construction on highway slopes, the specific steps of which are as follows:

[0009] Step 1: Collect data on vegetation coverage / lawn survival rate on highway slopes:

[0010] On the existing highway route map, some slope locations were randomly selected for on-site investigation of slope vegetation coverage or turf survival rate. For slopes that could be approached, the vegetation coverage / turf survival rate was estimated visually. At the same time, the slope location, slope protection method, slope aspect, slope position, altitude, slope gradient, backfill condition, and drainage facility settings were observed and recorded. For slopes that were difficult to observe, UAV aerial photography modeling was used to read vegetation coverage, basic engineering condition indicators, and basic plant growth construction indicators from the model map.

[0011] Methods for estimating vegetation cover or turf survival rate: Three observation plots are selected for each observation point, and a 1*1 vegetation quadrat is randomly set up in each plot. The vegetation cover (coverage in the plant quadrat survey) is estimated, and the average of the observation values ​​of the three observation plots is taken as the vegetation cover of the observation point; or 10 turf plots are randomly selected in each plot, and the proportion of surviving vegetation area on the turf plots is counted to the total area of ​​the turf plots. The average of the observation values ​​of the three observation plots is taken as the turf survival rate of the observation point.

[0012] Calibration of vegetation coverage or turf survival rate: Camera photos and drone photos were collected at slope observation points. Threshold segmentation method was applied on an indoor computer to extract the area of ​​vegetation coverage or turf survival rate in each observation plot. The vegetation coverage or turf survival rate was calculated by dividing the area by the area of ​​the observation plot. The final vegetation coverage or turf survival rate was obtained by combining the estimated values ​​on site and the threshold segmentation method calculated by the indoor computer.

[0013] All vegetation cover or lawn survival rate data from the survey points were compiled into one place. Using the Jenks' natural break method, the vegetation restoration effect of the slope was divided into three levels: 0-30% vegetation cover or lawn survival rate values ​​from low to high, indicating poor restoration effect; 30%-60% restoration effect; and 60%-90% restoration effect. The Jenks' natural break method is a commonly used classification method in this field.

[0014] Step 2: Investigation of basic engineering condition indicators and basic plant growth construction indicators:

[0015] At the sampling points where vegetation cover or lawn survival rate is investigated, basic engineering condition information is collected, including: slope location, slope protection type, slope aspect, slope position, slope gradient, and altitude.

[0016] Based on slope location, slopes are classified as: cut slopes and embankment roadbed slopes.

[0017] Slope protection methods include: arched grid soil stabilization and grass planting, soil slope three-dimensional mesh (or no) grass planting, rock slope topsoil spraying and grass planting, rectangular beam planting bags and grass planting, frame ecological bag grass planting, herringbone grid grass planting, hollow brick grass planting, etc., turf laying, arched frame + turf laying slope protection, frame beam + turf laying slope protection, etc. The choice depends on the site conditions or by referring to the design drawings.

[0018] Slope aspect and slope parameters were measured using a compass.

[0019] Slopes are divided into uphill slopes and downhill slopes;

[0020] Altitude was measured using GPS.

[0021] Basic infrastructure indicators for plant growth include: whether soil has been laid and whether drainage facilities have been installed. All of these indicators are derived from a combination of engineering design drawings and on-site observations.

[0022] Step 3: Establish a prediction model for slope vegetation coverage or lawn survival rate based on machine learning algorithms.

[0023] Based on the survey data collected in steps 1 and 2, a prediction model for the vegetation coverage or lawn survival rate of highway slopes was established using the random forest model in the machine learning algorithm, with the slope turf coverage or lawn survival rate as the predictor variable and the basic engineering condition index and the plant growth basic construction index as the explanatory variables.

[0024] Based on the variable categories, appropriate correlation analysis methods were selected to test the correlation between explanatory variables and predictor variables. Pearson correlation analysis was used for continuous variables, and Spearman correlation analysis was used for ordered categorical variables. Factors with significant correlation (P<0.1) with turf survival rate were selected as explanatory variables for the slope turf vegetation coverage or lawn survival rate prediction model.

[0025] Specifically, Pearson correlation analysis was used to analyze the correlation between altitude, slope and vegetation cover or turf survival rate. Spearman correlation analysis was used to analyze the relationship between slope type, protection type, slope aspect, slope position, backfill condition, drainage facility setting and vegetation cover or turf survival rate. Explanatory variables with significant correlation were selected (P<0.1). After calculation, except for the slope position variable, the correlation coefficients of vegetation cover or turf survival rate with other explanatory variables were all significantly less than 0.1, and could be used as explanatory variables for slope vegetation cover or turf survival rate.

[0026] Data on the survival rate of vegetation cover on highway slopes, along with explanatory variable data, were entered into a statistical software package in .CSV format for data cleaning and organization. This included defining missing values, defining variable formats, setting ordinal categorical variables as factor variables, setting continuous variables as integer variables, and setting the survival rate data of turf laying on highway slopes as integer variables. The `set.seed` function was used to generate a random number seed, the `randomForest` package was called, and the `creatDataPartion` command was used to generate a random training set list.

[0027] Furthermore, the field survey data is divided into a training set and a test set using a training set list. The training set accounts for 80% of the survey data, and the test set accounts for 20%. Using the training set as the model input data, the randomForest software package is called to generate a random forest model. The number of trees is checked to ensure it meets the requirements, and the regression effect of the model is determined. The importance function is called to calculate the InMSE (relative importance) and IncNodePurity (node ​​purity) indices, and the importance of each explanatory variable in the prediction model is ranked to determine the main influencing factors of slope turf vegetation coverage or lawn survival rate.

[0028] Step 4: Accuracy verification and optimization of the prediction model for vegetation cover or lawn survival rate on highway slopes, including:

[0029] Select the test set as input data, substitute it into the prediction model formed in step 3, call the predict function to generate predicted values ​​based on the test set, compare the roadside turf vegetation coverage or lawn survival rate in the test set with the predicted values ​​of roadside turf laying survival rate generated by the machine learning-based prediction model, evaluate the model prediction performance, and feed it back to the prediction model for variable optimization and variable selection, so that the prediction accuracy reaches more than 80%.

[0030] Step 5 involves decision-making regarding the establishment of a foundation for slope vegetation restoration and growth, including:

[0031] The basic engineering condition indicators and basic plant growth construction indicators extracted from the vegetation design documents of the proposed highway slope are used as explanatory variables. These are substituted into the verified and optimized prediction model formed in steps 3 and 4 to generate prediction results for the vegetation coverage or lawn survival rate of the proposed highway slope. The explanatory variables are ranked according to their importance, and optimization is performed sequentially. The decision indicators are iteratively optimized until the slope vegetation coverage or lawn survival rate reaches a level of good or satisfactory restoration. For explanatory variables that cannot be optimized, such as slope aspect, slope location, slope position, and altitude, optimization schemes involving engineering protection and changes in vegetation restoration measures are proposed.

[0032] Furthermore, camera photos and drone photos were collected at slope observation points. Threshold segmentation was applied on an indoor computer to extract the area of ​​vegetation cover or lawn survival rate within each observation plot. This area was then divided by the area of ​​the observation plot to calculate the vegetation cover or lawn survival rate. Combining the on-site estimated values ​​and the indoor computer threshold segmentation calculation, the final vegetation cover or lawn survival rate was obtained. This process included the following steps:

[0033] The extraction of vegetation coverage information from camera photos and drone photos of slope turf observation points mainly involves two steps: first, calculate the vegetation index, and then set an appropriate threshold to classify pixels with a vegetation index greater than the threshold as vegetation and pixels with a vegetation index less than the threshold as non-vegetation.

[0034] (1) The vegetation index is calculated as the visible light band difference vegetation index; the specific formula is as follows:

[0035]

[0036] Where, ρ green- Reflectivity and ρ in the green light band red- Reflectivity and ρ in the red band blue- Reflectivity in the blue light band;

[0037] (2) Determination of threshold

[0038] The thresholds for each vegetation index were determined using the bimodal histogram method and the histogram entropy threshold method, respectively. The extraction accuracy of the thresholds obtained by the two methods was compared, and the threshold with higher extraction accuracy was determined as the final threshold.

[0039] a. Bimodal histogram threshold determination method

[0040] The image contains a histogram with two obvious peaks. These two peaks correspond to a large number of points inside and outside the object, respectively. The valley between the two peaks corresponds to a relatively small number of target points near the edge of the object. That is, the valley is where the threshold is selected.

[0041] b. Histogram Entropy Threshold Method

[0042] First, assuming the threshold is t, the threshold t divides the image into two major categories: target O and background B. The entropy of the target region is HO(t), and the entropy of the background region is HB(t). The value of t that makes the total entropy H(t) = HO(t) + HB(t) reach its maximum value is the optimal threshold.

[0043] The thresholds for each vegetation index were determined using the bimodal histogram thresholding method and the histogram entropy thresholding method. The image was divided into vegetated and non-vegetated areas pixel by pixel using a human-computer interaction method. The results showed that VDVI had the highest overall accuracy in extracting vegetation regardless of whether the bimodal thresholding method or the histogram entropy thresholding method was used. Furthermore, the threshold extraction accuracy determined by the bimodal thresholding method was higher than that of the histogram entropy thresholding method.

[0044] Furthermore, this application also relates to the aforementioned prediction system for slope vegetation restoration and growth, including a data collection module: collecting data on vegetation coverage / lawn survival rate of highway slopes;

[0045] Data processing module: A module for predicting vegetation coverage or lawn survival rate on highway slopes;

[0046] Data output module: Generates prediction results of vegetation coverage or lawn survival rate of the proposed highway slope, sorted according to the importance of explanatory variables.

[0047] Furthermore, this application also relates to a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0048] Furthermore, this application also relates to a computer-readable storage medium storing a program that, when executed, can implement the above-described method.

[0049] The beneficial effects of this application are:

[0050] 1. The method for constructing a foundation for vegetation restoration and growth on highway slopes based on field surveys and machine learning proposed in this application involves collecting field data on vegetation restoration of existing highway slopes, constructing a predictive model, and incorporating the ecological restoration design of the proposed highway slope into the predictive model. This constructs a closed loop of data input, effect feedback, model optimization, and effect improvement feedback, thereby enhancing the ecological restoration effect of the proposed highway slope and forming a general method for constructing a foundation for vegetation restoration and growth on highway slopes.

[0051] 2. The method for constructing a growth foundation for slope vegetation restoration based on field surveys and machine learning proposed in this application can guide the optimization of the foundation engineering conditions of the proposed highway slope, guide the supplementation of the construction indicators for the plant growth foundation, thereby guiding the selection of slope vegetation restoration measures under different foundation engineering conditions, and optimizing the construction indicators for the plant growth foundation under different foundation engineering conditions, effectively improving the utilization efficiency of stripped turf in highway construction and the application effect of slope ecological restoration.

[0052] 3. This application's method, through a combination of field surveys and machine learning, establishes a general model for predicting slope turf planting. The model optimizes the indicators sequentially based on their importance, ranking them according to basic engineering conditions and vegetation growth foundation construction indicators, predicting the vegetation coverage or turf survival rate of the optimized slope until a good or relatively good level is achieved. For non-optimizable slope indicators such as slope aspect, slope location, slope position, and elevation, optimized slope restoration techniques are proposed.

[0053] 4. The method proposed in this application overcomes the shortcomings of existing engineering practices, such as the inability of highway slope foundation engineering design conditions to adapt to the ecological effects of slope turf laying, the lack of targeted plant growth foundation construction measures, and the inability to guarantee the restoration effect of slope turf laying. By using model prediction and variable optimization, the selection of highway slope turf laying technology is made more in line with the slope foundation engineering design conditions. With the assistance of plant growth foundation construction technology measures, the vegetation coverage / survival rate of the slope is improved, thereby providing support for the optimization of engineering design indicators and plant growth foundation construction decisions, and improving the ecological restoration effect of slope turf laying.

[0054] 5. This application overcomes the shortcomings of traditional methods, such as the large workload of observing the survival rate of turf on highway slopes, the difficulty in extending the observation results to other roads, and the lack of high-precision simulation and prediction models. It improves the targeting of ecological restoration measures for proposed highway slopes, enhances the utilization efficiency of turf stripping during highway construction, and improves the ecological restoration application effect, resulting in significant environmental and economic benefits.

[0055] 6. This application, through a creative comparative experimental study, identified eight influencing indicators affecting vegetation coverage or turf survival rate on highway slopes. Pearson correlation analysis was used to analyze the correlation between altitude, slope, and vegetation coverage or turf survival rate. Spearman correlation analysis was used to analyze the relationship between slope type, protection type, aspect, slope position, backfill condition, drainage facility setup, and vegetation coverage or turf survival rate. By screening factors with significant correlations to turf survival rate and ranking their importance, experiments showed that, except for slope position, the correlation coefficients between vegetation coverage or turf survival rate and other explanatory variables were all significantly less than 0.1, and therefore could be used as explanatory variables for slope vegetation coverage or turf survival rate. This resulted in a more optimized and accurate prediction model. Attached Figure Description

[0056] Figure 1 Flowchart of a method for predicting the survival rate of turf planting on highway slopes

[0057] Figure 2 Survival rate classification chart of turf planting on highway slopes

[0058] Figure 3 Relationship between random forest model error and number of trees

[0059] Figure 4 Ranking of the importance of factors affecting the survival rate of turf planting on highway slopes

[0060] Figure 5 Comparison chart of predicted and measured turf survival rates from the model in this application.

[0061] Figure 6 Comparison chart of predicted and measured values ​​based on random forest model

[0062] Figure 7 Image showing the improvement in slope vegetation coverage or lawn survival rate after optimization measures. Detailed Implementation

[0063] A proposed highway is located in the high-altitude and cold region of the Qinghai-Tibet Plateau, where the vegetation along the route is mainly alpine meadow. During the highway construction, in order to optimize the selection of slope turf planting techniques and the planting process, the method proposed by the applicant was adopted, as detailed below:

[0064] To more clearly illustrate this application, the following description, in conjunction with preferred embodiments and accompanying drawings, further clarifies the application. Similar components in the drawings are indicated by the same reference numerals. Those skilled in the art should understand that the specific description below is illustrative rather than restrictive and should not be construed as limiting the scope of protection of this application.

[0065] An embodiment of a method for vegetation restoration and construction on highway slopes includes the following steps:

[0066] Step 1: Collect data on vegetation coverage or lawn survival rate on highway slopes:

[0067] Along an existing highway in the Qinghai-Tibet Plateau region, randomly selected survey points were used to investigate the survival rate of turfgrass on slopes. For slopes that could be approached, visual estimation was used to estimate vegetation cover or turf survival rate. Simultaneously, slope location, slope protection method, aspect, slope position, altitude, slope gradient, backfill condition, and drainage facility installation were observed and recorded. For slopes that were difficult to observe, drone aerial photography and modeling were used to read vegetation cover or turf survival rate, as well as basic engineering conditions and plant growth infrastructure indicators from the model map.

[0068] Methods for estimating vegetation cover / survival rate: Three observation plots are selected for each observation point, and a 1*1 vegetation quadrat is randomly set up in each plot. The vegetation cover (coverage in the plant quadrat survey) is estimated, and the average of the observation values ​​of the three observation plots is taken as the vegetation cover / survival rate of the observation point; or 10 turf plots are randomly selected in each plot, and the proportion of surviving vegetation area on the turf plots is counted to the total area of ​​the turf plots. The average of the observation values ​​of the three observation plots is taken as the vegetation cover / survival rate of the observation point.

[0069] Slope turf survival rate calibration: Digital camera and drone photos were collected at slope turf observation points. Threshold segmentation was applied on an indoor computer to extract the surviving area of ​​turf in each observation plot, calculating its percentage of the total plot area. The final turf survival rate was obtained through rapid on-site observation and indoor computer calibration. A total of 172 points were investigated to assess turf survival rates and influencing factors.

[0070] All lawn survival rate data from the survey sites were compiled and analyzed using Jenks' natural break method to classify slope vegetation restoration into three levels: 0-30% (poor restoration), 30%-60% (relatively good restoration), and 60%-90% (excellent restoration). See details below. Figure 2 .

[0071] Furthermore, the extraction of vegetation coverage information from the camera photos and drone photos of the slope turf observation points in step 1 mainly consists of two steps: first, calculate the vegetation index, and then set an appropriate threshold to classify pixels with a vegetation index greater than the threshold as vegetation and pixels with a vegetation index less than the threshold as non-vegetation.

[0072] (1) The vegetation index is calculated as the visible light band difference vegetation index; the specific formula is as follows:

[0073]

[0074] Where, ρ green- Reflectivity and ρ in the green light band red- Reflectivity and ρ in the red band blue- Reflectivity in the blue light band;

[0075] (2) Determination of threshold

[0076] The thresholds for each vegetation index were determined using the bimodal histogram method and the histogram entropy threshold method, respectively. The extraction accuracy of the thresholds obtained by the two methods was compared, and the threshold with higher extraction accuracy was determined as the final threshold.

[0077] a. Bimodal histogram threshold determination method

[0078] The image contains a histogram with two obvious peaks. These two peaks correspond to a large number of points inside and outside the object, respectively. The valley between the two peaks corresponds to a relatively small number of target points near the edge of the object. That is, the valley is where the threshold is selected.

[0079] b. Histogram Entropy Threshold Method

[0080] First, assuming the threshold is t, the threshold t divides the image into two major categories: target O and background B. The entropy of the target region is HO(t), and the entropy of the background region is HB(t). The value of t that makes the total entropy H(t) = HO(t) + HB(t) reach its maximum value is the optimal threshold.

[0081] Thresholds for each vegetation index were determined using the bimodal histogram thresholding method and the histogram entropy thresholding method, as shown in the table below. Vegetation information was extracted using the thresholds determined by these two methods, yielding the vegetation distribution results corresponding to each vegetation index. Furthermore, an interactive method was used to divide the image pixel by pixel into vegetated and non-vegetated areas.

[0082] Table 1 Vegetation Index and Thresholds

[0083] Vegetation Index Bimodal histogram thresholding method Histogram entropy thresholding method VDVI 0.039217 0.0743165

[0084] The accuracy evaluation of vegetation regions extracted by each vegetation index is shown in the table below. It shows that regardless of whether the bimodal thresholding method or the histogram entropy thresholding method is used, VDVI has the highest overall accuracy in extracting vegetation, and the threshold determined by the bimodal thresholding method has higher extraction accuracy than that determined by the histogram entropy thresholding method.

[0085] Table 2 Evaluation Table of Vegetation Extraction Accuracy

[0086] vegetation non-vegetation Overall accuracy Kappa coefficient Bimodal threshold method 98.26 99.15 98.56 0.97 Histogram entropy thresholding method 86.88 99.81 91.29 0.82

[0087] Step 2: Investigation of basic engineering condition indicators and basic plant growth construction indicators:

[0088] At the sampling points investigating lawn survival rates, basic engineering condition information was collected, including: slope location, slope protection method, slope aspect, slope position, slope gradient, and elevation; among which:

[0089] Based on the location of the slope, it is divided into: cut slope and embankment roadbed slope;

[0090] Slope protection methods include: arched grid soil stabilization and grass planting, soil slope three-dimensional mesh (or no) grass planting, rock slope topsoil spraying and grass planting, rectangular beam planting bags and grass planting, frame ecological bag grass planting, herringbone grid grass planting, hollow brick grass planting, etc., turf laying, arched frame + turf laying slope protection, frame beam + turf laying slope protection, etc. The choice depends on the site conditions and the design drawings.

[0091] Slope aspect and slope parameters were measured using a compass.

[0092] Slope: Uphill, downhill;

[0093] Altitude was measured using GPS.

[0094] Basic construction indicators for plant growth include: whether soil is laid and whether drainage facilities are installed. These indicators are obtained by combining engineering design drawings with on-site observations.

[0095] Step 3: Establish a slope turf survival rate prediction model based on machine learning algorithms;

[0096] Based on the survey data collected in steps 1 and 2, a prediction model for the survival rate of turf on highway slopes was established using the random forest model in the machine learning algorithm, with slope vegetation coverage or lawn survival rate as the predictor variable and basic engineering condition indicators and plant growth basic construction indicators as explanatory variables.

[0097] Pearson correlation analysis was used to analyze the correlation between elevation and slope as predictors and turf survival rate, as shown in Table 1. Spearman correlation analysis was used to analyze the relationship between ordered categorical predictors such as slope type, protection type, aspect, slope position, backfill condition, and drainage facility setting and turf survival rate, as shown in Table 3. Variables with significant correlation to turf survival rate (P<0.1) were selected. After calculation, except for the slope position factor, the correlation coefficients of other explanatory variables were all less than 0.1, and they can be used as explanatory variables in the slope turf survival impact prediction model.

[0098] Table 3. Factors affecting turf survival rate

[0099]

[0100] The prediction model based on random forest was constructed by collecting data on the survival rate of turf on slopes and its influencing factors from all observation points. 80% of the samples (140 trees) were randomly selected as the training samples for the random forest model. Eight influencing factors were used as explanatory variables, and the survival rate of turf patches was used as the predictor variable. The Randomforest toolkit in RStudio software was used to simulate and generate random trees. First, the relationship between model error and the number of random trees was determined, and the maximum number of random trees was determined. In this example, when the number of random trees ntree = 500, the error basically stabilized. See [link to documentation]. Figure 3 The remaining 20% ​​of the data will be used as test data for subsequent predictive analysis.

[0101] The importance ranking of influencing factors uses the `importance` function to calculate two indices: `InMSE` (relative importance) and `IncNodePurity` (node ​​purity), determining the importance of each independent variable in the random forest model. Figure 4 As shown, this illustrates the importance of factors affecting slope turf survival. These are, in order: slope aspect, elevation, slope gradient, slope protection method, slope location, backfill condition, drainage facilities, and slope position.

[0102] Step 4: Accuracy verification and optimization of the highway slope lawn survival rate prediction model, including:

[0103] The generated random forest prediction model for slope turf patch survival rate was used to predict the survival rate of 20% of the test data (32 patches). The predicted turf patch survival rate was obtained based on the index values ​​of each explanatory variable, as shown in Table 4. The prediction results and actual observed survival rate data were visualized using scatter plots and line graphs. The model's prediction accuracy reached 83.28%, as shown in Table 4. Figure 5 As shown, the overall prediction results generally match the actual turf patch survival rate in the test data well. This indicates that the random forest model used can effectively predict the survival rate of turf patches on slopes.

[0104] Table 4. Measured and predicted survival rates of turf planting on highway slopes.

[0105]

[0106]

[0107] Step 5 involves decision-making regarding the establishment of a foundation for slope vegetation restoration and growth, including:

[0108] The prediction model formed in step 4 was used in the design of a proposed highway slope restoration project in the Qinghai-Tibet Plateau region. The basic engineering conditions and basic vegetation growth indicators of the slope restoration project were substituted into the prediction model to predict the survival rate of the turf. The predictions for the proposed highway slope are shown in Table 5.

[0109] Table 5. Optimization of Indicators and Predicted Effects of the Proposed Highway Slope Repair Project

[0110]

[0111]

[0112] Based on the importance ranking of the predictor variables, the optimizable indicators are optimized in the following order: slope protection method, slope aspect, slope gradient, soil filling condition, and drainage facility setting. These are then substituted into the prediction model to predict the survival rate of the grass on the optimized slope.

[0113] After implementing optimized protection measures, the average survival rate of the slope turf increased by 17.13%; after optimizing the slope gradient, the average increase was 16.00%; after adding soil filling, the average increase was 18.57%; and after implementing 2-3 of the above optimization measures, the average increase was 32.08%. After optimization, the turf survival rate at all 14 slopes reached a good or relatively good level. Figure 6 ).

[0114] The method described in this application forms a foundation for the restoration and growth of vegetation on highway slopes, which can optimize the greening methods for highway slopes, select suitable ecological restoration methods, and improve the greening effect of slopes.

[0115] It should be understood that the above detailed description of the technical solutions of this application by means of preferred embodiments is illustrative and not restrictive. Those skilled in the art can modify the technical solutions described in the embodiments or make equivalent substitutions for some of the technical features based on reading this application specification; however, such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for establishing and constructing a foundation for vegetation restoration on highway slopes, characterized in that, Based on surveys of vegetation coverage, basic engineering conditions, and basic vegetation growth indicators on existing highway slopes, machine learning methods were used to establish prediction models for slope vegetation coverage under different scenarios. The proposed ecological restoration design scheme for the highway slopes was then substituted into the prediction models. The indicators were optimized sequentially according to their importance, prioritizing basic engineering conditions and vegetation growth indicators. This process of optimization-prediction-re-optimization-re-prediction was continuously refined until the slope vegetation restoration effect reached a good or relatively good level. For slope aspect, location, position, and elevation indicators that could not be optimized, optimized slope restoration techniques were proposed, with the specific steps as follows: Step 1: Collect vegetation coverage data of highway slopes; obtain vegetation coverage data of existing highway slopes through field surveys; Step 2: Investigation of basic engineering condition indicators and basic construction indicators for plant growth; the basic engineering condition indicators include: slope location, slope protection type, slope aspect, slope position, slope gradient, and altitude; the basic construction indicators for plant growth include: whether soil is laid and whether drainage facilities are installed. Step 3: Establish a slope vegetation coverage prediction model based on machine learning algorithms; specifically, based on the survey data collected in Step 1 and Step 2, use machine learning algorithms to establish a highway slope vegetation coverage or lawn survival rate prediction model with slope vegetation coverage or lawn survival rate as the predictor variable and basic engineering condition indicators and plant growth basic construction indicators as explanatory variables. Step 4: Accuracy verification and optimization of the highway slope vegetation coverage prediction model; Based on the prediction model formed in Step 3, compare and evaluate the predicted values ​​with the field observation values; Step 5: Slope vegetation restoration, growth and construction. The basic engineering condition indicators and basic plant growth construction indicators extracted from the ecological restoration design documents of the proposed highway slope are used as explanatory variables and substituted into the prediction model to generate the prediction results of the vegetation coverage or lawn survival rate of the proposed highway slope. The explanatory variables are optimized in order of importance according to their importance. Through iteration, the vegetation coverage or lawn survival rate of the slope is gradually improved to a better or good level.

2. The method as described in claim 1, characterized in that, Step 1 specifically involves randomly selecting some points on the highway route map to investigate the vegetation coverage or lawn survival rate data of the slopes. For slopes that can be approached, the vegetation coverage is estimated by visual estimation. At the same time, the slope location, slope protection form, slope aspect, slope position, altitude, slope gradient, backfill condition, and drainage facility settings are observed and recorded. For slopes that are difficult to approach and observe due to traffic reasons, the drone aerial photography modeling method is used to read the vegetation coverage or lawn survival rate on the model map. Calibration of vegetation coverage or lawn survival rate: Collect digital camera photos and drone photos of slope observation points, apply threshold segmentation method on indoor computer to extract the area of ​​vegetation coverage or lawn survival rate in each observation plot, divide by the area of ​​the observation plot, calculate vegetation coverage or lawn survival rate, and obtain the final vegetation coverage or lawn survival rate through rapid on-site observation and indoor computer calibration. The vegetation coverage or lawn survival rate data of all survey points were compiled into a data table. Using the Jenks natural break method, the vegetation restoration effect of the slope was divided into three levels: vegetation coverage or lawn survival rate values ​​from low to high. 0-30% indicates poor restoration effect, 30%-60% indicates relatively good restoration effect, and 60%-90% indicates good restoration effect.

3. The method as described in claim 1, characterized in that, Step 2 specifically involves collecting basic engineering condition indicators at sampling points used to investigate vegetation cover or lawn survival rate, including: slope location, slope protection method, slope aspect, slope position, slope gradient, and altitude; among which, The slope locations are divided into: cut slopes and fill roadbed slopes; Slope protection methods are divided into several types: arched grid soil stabilization and grass planting, three-dimensional mesh grass planting for soil slopes, topsoil spraying and grass planting for rock slopes, rectangular beam planting bags for grass planting, frame ecological bags for grass planting, herringbone grid grass planting, hollow brick grass planting, turf laying, arched frame + turf laying slope protection, and frame beam + turf laying slope protection. Slope aspect and slope parameters were measured using a compass. Slope: Divided into uphill slope and downhill slope; Altitude was measured using GPS. Basic indicators for plant growth include: whether soil is laid and whether drainage facilities are installed. These indicators are obtained by combining engineering design drawings or on-site observations.

4. The method as described in claim 1, characterized in that, Step 3 includes the following steps: Step 3.1: Based on the survey data collected in Steps 1 and 2, a prediction model for the vegetation coverage or lawn survival rate of highway slopes is established using the random forest model in the machine learning algorithm, with slope vegetation coverage or lawn survival rate as the predictor variable and basic engineering condition indicators and plant growth basic construction indicators as explanatory variables. Step 3.2: Based on the categories of ordinal and continuous variables, select the appropriate correlation analysis method to test the correlation, and screen the factors that have a significant correlation with slope vegetation coverage or lawn survival rate (P<0.1) as explanatory variables for the vegetation coverage or lawn survival rate prediction model. Step 3.3: Input the highway slope survival rate data and influencing factor data into the machine learning software package in .CSV data table or matrix format. After data cleaning and organization, including defining missing values, defining variable formats, setting ordered categorical variables as factor variables, continuous variables as integer variables, and setting slope vegetation coverage or lawn survival rate data as integer variables, a data structure usable for machine learning is formed. Use the set.seed function to generate random number seeds, call the randomForest package, and use the createDataPartion command to generate a random training set list. Step 3.4: Divide the survey data into a training set and a test set using the training set list. The training set data accounts for 80% of the field survey data, and the test set accounts for 20%. Call the applicable machine learning algorithm to build a prediction model. Further determine the importance of variables in the model and rank them to identify the main influencing factors.

5. The method as described in claim 4, characterized in that, Step 3.2 specifically involves using Pearson correlation analysis to analyze the correlation between altitude, slope, and vegetation cover or turf survival rate. Spearman correlation analysis is used to analyze the relationship between slope type, protection type, slope aspect, slope position, backfill condition, drainage facility setting, and vegetation cover or turf survival rate. Explanatory variables with significant correlations (P < 0.1) are selected. After calculation, except for the slope position variable, the correlation coefficients of vegetation cover or turf survival rate with other explanatory variables are all significantly less than 0.1, and can be used as explanatory variables for slope vegetation cover or turf survival rate.

6. The method as described in claim 4, characterized in that, Step 3.4 specifically involves constructing a prediction model based on random forest. By organizing the data on slope vegetation coverage or turf survival rate and its influencing factors at all observation points, 80% of the samples are randomly selected as training samples for the random forest model. Eight influencing factor indicators are used as explanatory variables, and vegetation coverage or turf survival rate is used as predictive variables. Random trees are generated using the Randomforest toolkit in RStudio software. First, the relationship between model error and the number of random trees is determined, and the maximum number of random trees is determined. The remaining 20% ​​of the data is used as test data for subsequent prediction analysis. The importance of influencing factors is ranked by calling the importance function to calculate the InMSE (relative importance) and IncNodePurity (node ​​purity) in the slope vegetation coverage or turf survival prediction model.

7. The method as described in claim 1, characterized in that, Step 4 specifically involves selecting a test set as input data, substituting it into the prediction model formed in step 3, calling the predict function to generate predicted values ​​based on the test set, comparing the roadside turf vegetation coverage or lawn survival rate in the test set with the predicted values ​​of roadside turf laying survival rate generated by the machine learning-based prediction model, evaluating the model's prediction performance, and feeding it back to the prediction model for variable optimization and selection, so that the prediction accuracy reaches over 80%.

8. The method as described in claim 1, characterized in that, Step 5 specifically involves taking the basic engineering condition indicators and plant growth basic construction indicators extracted from the vegetation design documents of the proposed highway slope as explanatory variables, substituting them into the verified and optimized prediction model formed in steps 3 and 4, generating prediction results for the vegetation coverage or lawn survival rate of the proposed highway slope, optimizing them sequentially according to the importance of the explanatory variables, and gradually improving the restoration effect of slope vegetation coverage or lawn survival rate to a good or good level through iteration. For non-optimizable indicators among the explanatory variables, such as slope aspect, slope location, and altitude, optimization schemes are proposed by adopting engineering protection and vegetation restoration measures.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed, enables the implementation of the method as described in any one of claims 1-8.

Images