Slope ecological restoration method, device, electronic equipment and readable medium
By optimizing slope ecological restoration schemes using deep learning models and the analytic hierarchy process, the problems of poor restoration results and high costs caused by expert scoring methods were solved, and a more reliable restoration scheme selection was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 中国地质环境监测院(自然资源部地质灾害技术指导中心)
- Filing Date
- 2025-12-09
- Publication Date
- 2026-06-09
Smart Images
Figure CN122175750A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of ecological restoration technology, and more specifically, to a method, apparatus, electronic device, and readable medium for ecological restoration of slopes. Background Technology
[0002] In related technologies, slope ecological restoration mainly revolves around two objectives: ensuring slope stability and creating vegetation habitats. Slope stability is primarily achieved using plant-based soil nailing, vegetated concrete, and three-dimensional vegetation netting techniques; vegetation habitat creation mainly utilizes hole planting, fish-scale pit planting, and floating platform greening techniques. When performing ecological restoration of slopes, the appropriate restoration technology is selected based on the actual conditions of the slope. However, in determining the appropriate restoration scheme, expert scoring is often used to assess the restoration effects of different schemes and select the optimal one. This method, being a manual evaluation, is subjective and may result in low reliability of the optimal restoration scheme, potentially leading to poor restoration effects and high costs. Summary of the Invention
[0003] The main purpose of this application is to provide a slope ecological restoration method, device, electronic device and readable medium to solve the problem that the selection of restoration schemes in existing slope ecological restoration methods is carried out by manual evaluation, which may lead to poor restoration effect and high restoration cost.
[0004] According to one aspect of this application, a slope ecological restoration method is provided, comprising: Step S1: Determine the ecological restoration scheme for the slope to be restored, generate a slope matrix based on the ecological restoration scheme, train a deep learning model based on the slope matrix, and obtain a trained decoder. Step S2: Construct an evaluation model, determine the weights of evaluation indicators using the analytic hierarchy process, determine the evaluation score of the slope ecological restoration scheme based on the evaluation indicator weights, and optimize the evaluation model according to the slope matrix and the evaluation score to obtain the optimized evaluation model; Step S3: The initial slope matrix is generated iteratively through the decoder, and the initial slope matrix is evaluated through the optimized evaluation model. Based on the evaluation results, the optimal slope matrix is obtained. Step S4: Decode the optimal slope matrix to determine the optimal ecological restoration scheme for the slope to be restored, and restore the slope to be restored according to the optimal ecological restoration scheme.
[0005] Further, step S1 includes: Step S11: Obtain the repair category, slope angle, and slope type corresponding to the slope to be repaired; Step S12: Based on the repair category, the slope angle, and the slope type, obtain the slope ecological restoration plan corresponding to the slope to be repaired through a preset initial mapping relationship; The initial mapping relationship is configured as the correspondence between the restoration category, the slope angle, the slope type, and the slope ecological restoration scheme; Step S13: Construct an initial coding matrix based on the slope ecological restoration scheme. Fill in the initial coding matrix to obtain the slope matrix according to the correspondence between the slope angle, the slope type, the restoration category and the slope ecological restoration scheme. Step S14: Construct a deep learning model based on an encoder-decoder structure, train the deep learning model based on historical data, and remove the decoder module from the trained deep learning model to obtain the trained decoder.
[0006] Furthermore, The restoration categories include Class I restoration and Class II restoration. Class I restoration is configured to improve the stability of the slope to be restored, and Class II restoration is configured to create habitats for the slope to be restored; and / or, The slope types include at least soil slopes, weakly weathered rock slopes, and hard rock slopes; and / or, The slope ecological restoration plan includes at least the following: plant nailing plan, vegetation concrete plan, three-dimensional vegetation net plan, hole planting plan, fish scale pit plan, and ledge greening plan.
[0007] Furthermore, the evaluation model is configured as a multilayer perceptron model, and the model structure corresponding to the evaluation model includes an input layer, an output layer, a first hidden layer, and a second hidden layer. The input layer is configured with the same dimension as the slope matrix; the first hidden layer includes 64 neurons, the second hidden layer includes 32 neurons, and the activation functions of both the first and second hidden layers are ReLU functions; the output layer includes 1 neuron for outputting the evaluation score, and the activation function of the output layer is the Sigmoid function.
[0008] Further, step S2 includes: Step S21: Construct an evaluation model based on a multilayer perceptron; Step S22: Determine the evaluation index of the slope to be repaired, wherein the evaluation index includes primary indexes and at least one secondary index corresponding to each primary index; Step S23: Based on the evaluation indicators, analyze the historical data using the analytic hierarchy process (AHP) to determine the weight vector corresponding to each evaluation indicator, and calculate the evaluation score of the slope ecological restoration scheme based on the weight vector using the weighted summation method. Step S24: Use the slope matrix as the input sample and the evaluation score as the label to construct an optimization dataset; Step S25: Based on the optimized dataset, optimize the evaluation model using the mean squared error loss function and the Adam optimizer to obtain the optimized evaluation model.
[0009] Further, step S3 includes: Step S31: Construct an initial population based on the latent space dimension corresponding to the trained decoder; Step S32: Set an objective function with the goal of maximizing the evaluation score output by the evaluation model, and set the fitness function to map to the output score of the evaluation model; Step S33: Input each latent vector in the initial population into the trained decoder to obtain the initial slope matrix. In each iteration, generate multiple initial slope matrices to form a matrix set. Step S34: Input the set of matrices into the optimized evaluation model, output the evaluation score corresponding to each initial slope matrix through the first hidden layer and the second hidden layer, and sort them in descending order. The initial slope matrices with scores greater than the preset value and their corresponding hidden vectors are retained as high-quality individuals. Step S35: Retain a preset number of high-quality individuals, and randomly mutate the latent vectors corresponding to the remaining high-quality individuals to form a new population; Step S36: Using the new population as input, iteratively execute steps S33 to S35. When the fluctuation of the highest evaluation score corresponding to m consecutive rounds is less than or equal to the difference threshold or the number of iterations reaches n rounds, terminate the iteration and select the initial slope matrix with the highest evaluation score as the optimal slope matrix.
[0010] Further, step S4 includes: Step S41: Analyze the optimal slope matrix and extract the correspondence between the optimized restoration category, the slope angle, and the slope type and the ecological restoration scheme in the slope ecological restoration scheme; Step S42: Based on the optimized correspondence between the repair category, the slope angle, the slope type, and the slope ecological restoration scheme, rematch the slope ecological restoration scheme corresponding to the slope to be restored as the optimal slope ecological restoration scheme; Step S43: Set the final restoration plan for the slope to be restored according to the optimal slope ecological restoration plan, so as to realize the ecological restoration of the slope to be restored.
[0011] According to another aspect of this application, a slope ecological restoration device is provided for performing the above-described slope ecological restoration method, comprising: The decoder training module is used to determine the ecological restoration scheme corresponding to the slope to be restored, generate a slope matrix based on the ecological restoration scheme, and train a deep learning model based on the slope matrix to obtain a trained decoder. The model generation and optimization module is used to construct an evaluation model, determine the weights of evaluation indicators through the analytic hierarchy process, determine the evaluation score of the slope ecological restoration scheme based on the evaluation indicator weights, and optimize the evaluation model according to the slope matrix and the evaluation score to obtain the optimized evaluation model. An optimal matrix filtering module is used to iteratively generate an initial slope matrix through the decoder, evaluate the initial slope matrix through an optimized evaluation model, and obtain the optimal slope matrix based on the evaluation results. The repair scheme determination module is used to decode the optimal slope matrix, determine the optimal ecological restoration scheme for the slope to be restored, and restore the slope to be restored according to the optimal ecological restoration scheme.
[0012] According to another aspect of this application, an electronic device is provided, comprising: a processor, and a memory storing a program, the program including instructions that, when executed by the processor, cause the processor to perform the above-described slope ecological restoration method.
[0013] According to another aspect of this application, a non-transitory machine-readable medium storing computer instructions is provided, the computer instructions being used to cause the computer to perform the above-described slope ecological restoration method.
[0014] In this application, a slope matrix corresponding to the slope restoration scheme is generated iteratively by a decoder based on a deep learning model. The weights of different evaluation indicators are determined by the analytic hierarchy process (AHP) to optimize the evaluation model. The optimized evaluation model is then used to evaluate the slope matrix to determine the optimal slope matrix, thereby obtaining the optimal slope restoration scheme. This approach avoids the unreliability of restoration schemes caused by subjective evaluation methods such as expert scoring. It enables the optimal restoration scheme to be selected through iterative evaluation model selection, improving the ecological restoration effect of the slope and optimizing the restoration cost. Attached Figure Description
[0015] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, are illustrative and descriptive, serving to explain this application and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a schematic flowchart of a slope ecological restoration method disclosed in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a slope ecological restoration device disclosed in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. Detailed Implementation
[0016] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0017] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0018] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values of the components and steps set forth in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.
[0019] As described in the background section, in related technologies, slope ecological restoration mainly employs expert scoring to assess the restoration effect of corresponding restoration schemes and select the optimal scheme. However, this method is a manual assessment, which is subjective and results in low reliability of the optimal restoration scheme, potentially leading to poor restoration effects and high costs. Therefore, this application provides a slope ecological restoration method, apparatus, electronic equipment, and readable medium. The slope ecological restoration method generates a restoration scheme matrix using a deep learning model and optimizes the evaluation model using the analytic hierarchy process (AHP). The optimized evaluation model can more accurately determine the optimal restoration scheme, improving the effectiveness of slope ecological restoration. The following description, in conjunction with accompanying drawings, illustrates how this application uses a deep learning model to generate a restoration scheme matrix and AHP to optimize the evaluation model, objectively selecting the optimal restoration scheme and addressing the problems of poor restoration effects and high costs caused by the subjectivity of manual assessment.
[0020] See Figure 1 As shown in the embodiment of this application, a slope ecological restoration method is provided, including: Step S1: Determine the ecological restoration scheme for the slope to be restored, generate a slope matrix based on the ecological restoration scheme, train a deep learning model based on the slope matrix, and obtain a trained decoder.
[0021] Specifically, step S1 includes: Step S11: Obtain the repair category, slope angle, and slope type corresponding to the slope to be repaired; Step S12: Based on the repair category, slope angle, and slope type, obtain the slope ecological restoration plan corresponding to the slope to be repaired through the preset initial mapping relationship; The initial mapping relationship is configured as the correspondence between the restoration category, slope angle, and slope type and the slope ecological restoration scheme. Step S13: Construct an initial coding matrix based on the slope ecological restoration scheme. According to the correspondence between slope angle, slope type, and restoration category and slope ecological restoration scheme, fill in the initial coding matrix to obtain the slope matrix. Step S14: Construct a deep learning model based on an encoder-decoder structure, train the deep learning model based on historical data, and extract the decoder module from the trained deep learning model to obtain the trained decoder.
[0022] In this embodiment, by determining the repair category, slope angle, and slope type of the slope to be repaired, and combining this with a preset mapping relationship, an initial directional matching between the slope to be repaired and the repair scheme can be achieved, improving the targeted nature of the repair. Transforming the abstract repair scheme into a structured slope matrix and integrating the correlation logic between parameters and schemes provides a standardized data carrier for subsequent model processing, enabling quantitative analysis of slope ecological restoration schemes. By training an encoder-decoder model and then removing the decoder, the model can learn the correlation between parameters and repair schemes in historical data, improving its ability to optimize subsequent schemes.
[0023] In some embodiments, the repair categories corresponding to the slope to be repaired in this embodiment include Class I repair and Class II repair. Class I repair is configured to improve the stability of the slope to be repaired, and Class II repair is configured to create habitats for the slope to be repaired. It can be understood that Class I repair indicates that the main requirement for the slope to be repaired is to improve its stability, while Class II repair indicates that the main requirement for the slope to be repaired is to restore its habitat. Simultaneously, while improving the stability of the slope to be repaired, habitat creation and repair are also required, and vice versa.
[0024] In some embodiments, the slope types corresponding to the slope to be repaired in this embodiment include at least soil slopes, weakly weathered rock slopes, and hard rock slopes.
[0025] In some embodiments, the ecological restoration schemes for the slope to be restored in this embodiment include at least the following: plant nailing scheme, vegetation concrete scheme, three-dimensional vegetation net scheme, pit planting scheme, fish-scale pit scheme, and floating platform greening scheme. Among these, the plant nailing scheme, vegetation concrete scheme, and three-dimensional vegetation net scheme are mainly suitable for Class I restoration and can significantly improve the stability of the slope to be restored. The pit planting scheme, fish-scale pit scheme, and floating platform greening scheme are mainly suitable for Class II restoration and can reliably create and restore habitats for the slope to be restored.
[0026] Understandably, the plant-based soil nailing scheme in this embodiment is a slope stabilization method that combines ecological engineering and geotechnical technology. It is primarily used to improve slope stability while increasing the green area of the slope, promoting ecological restoration and improvement. Traditional soil nails are used to enhance the mechanical stability of the slope, and the root system of vegetation is combined to further stabilize the soil and improve the ecology.
[0027] Specifically, the process of a plant-based soil nailing scheme includes the installation of soil nails, selection of suitable plants, planting of plants, improvement of soil conditions, and restoration of ecology and function. First, biodegradable soil nails (such as wood or other natural materials) are inserted vertically or at an angle into the slope soil. The soil nails help physically stabilize the slope and provide a supporting structure for plant growth. Plants that are adaptable to the local environment of the slope to be restored, grow rapidly, and have well-developed root systems are selected; herbaceous or shrubby plants are typically used to ensure their roots further anchor the soil and reduce soil erosion. The selected plants are planted directly around the soil nails or in pre-drilled holes in the soil nails. During growth, the plant roots gradually penetrate and envelop the soil nails, enhancing soil cohesion. In this embodiment, soil improvement may be necessary concurrently with soil nail installation and planting, such as adding organic fertilizers and improving drainage conditions, to promote plant growth and improve the overall quality of the ecosystem. With plant growth and the stabilizing effect of the soil nails, the ecological function of the slope can gradually be restored. The planted plants not only improve the microclimate and provide habitats for biodiversity, but also increase carbon sequestration and reduce greenhouse gas emissions through photosynthesis.
[0028] Understandably, the vegetation concrete scheme in this embodiment primarily utilizes high-strength binders (concrete, cement, etc.) to firmly shape larger-diameter aggregates. The pores between the aggregates store a substrate (AB bacteria, organic matter, sandy loam, plant seeds, fertilizer, water, etc.) that supports plant growth. Through sowing or other methods, various plants grow within this robust aggregate concrete substrate layer, thus restoring the ecological environment through vegetation. This scheme uses cement, which enhances slope strength and erosion resistance, quickly creating a suitable environment for plant growth. It effectively solves the problems of rock slope protection and rapid greening. Furthermore, the combined effect of concrete, vegetation, and the substrate enhances the substrate's erosion resistance, effectively ensuring rapid vegetation formation and ecological stability. This technology boasts a high degree of mechanization and large production capacity. Construction utilizes a dry-spraying machine for spraying, achieving a long spraying distance. The sprayed layer possesses sufficient strength and is not prone to cracking, exhibiting strong erosion resistance.
[0029] Specifically, the vegetation concrete scheme involves slope treatment, installation of wire mesh and anchors, spraying of vegetation concrete, planting of vegetation concrete, base layer spraying, and maintenance spraying. First, all obstacles hindering construction on the slope must be cleared, including removing vegetation junctions, weeds, fallen leaves and branches, loose soil and stones, etc., as well as slope trimming. When clearing vegetation junctions, focus on cleaning the contact surface of the original slope above the slope opening line. The cleaning width can be set at 1.0–1.5m, aiming to remove the branches and stems of plants on the original slope. Underground roots do not need to be removed; this part serves as the transition between the project and the original slope, i.e., the vegetation junction. For protruding, easily detached areas with obvious unstable rocks, use an electric hammer or pneumatic drill to drill holes along the slope at the protruding area, and then use a hammer to knock them down. For areas with obvious depressions, use a pneumatic pick to roughen the surface of the area to be filled to a depth of 1 cm or more. Then, clean it with high-pressure air and water, and finally fill it with M7.5 mortar to ensure a good filling effect. Install anchors perpendicular to the slope according to the designed anchor specifications, rock penetration depth, and spacing. Drill holes perpendicular to the slope using an electric hammer and drive the anchors in. Increase the density and length of anchors at the top of the slope to enhance stability. The driven anchors should be slightly inclined upwards, forming an angle of 95-100° with the slope. In areas where the rock is severely weathered, the anchors should be lengthened as needed to ensure stability after being driven into the slope. Lay chain link wire mesh, starting from the top of the vegetation junction and working downwards. The reinforced mesh should be laid taut, with a minimum 5 cm overlap between vertical sections, but no horizontal overlap. All mesh panels should be securely tied to the anchors, maintaining a distance between the mesh panels and the slope surface; otherwise, use padding blocks for support. Prepare the components according to the mixing ratio, and mix them thoroughly using a mixer before use. Plant seeds as required by the design should be added during the mixing of the surface substrate. After completing slope treatment, netting and anchor installation, and preparing and mixing the components of the vegetation concrete substrate, the spraying of the vegetation concrete substrate can begin. The equipment used for spraying is a general concrete spraying machine, and the base layer and surface layer are treated separately. Spraying is carried out from top to bottom on the slope, first the base layer and then the surface layer, with each spraying section being 4-6 meters wide and 3-5 meters high. Dry spraying is used for spraying, generally powered by an air compressor with a capacity of more than 12 cubic meters. Before spraying the base layer, check the slope surface again to ensure that the loose soil, turf, tree roots and other debris are clean. After confirming that the surface is clean, spray the slope with water to promote a tight bond between the sprayed vegetation concrete substrate and the base surface. Then, conduct a test spraying to adjust the water-cement ratio before proceeding with the spraying construction. The thickness of the base layer spraying should be 8-9 cm. When starting the spraying operation, the air should be supplied first, then the machine should be turned on, and then the material should be fed. When the spraying is finished, the air should be turned off after all the sprayed material has been sprayed.The base layer shotcrete can be sprayed to the designed thickness in one go, without the need for layered spraying. During spraying, the distance between the nozzle and the slope surface should be controlled between 0.6m and 1.0m, and it should generally be perpendicular to the slope surface, with a maximum tilt angle not exceeding 10°. During spraying, the output pressure of the spray head should not be less than 0.1MPa. Spraying should be carried out from top to bottom, first spraying the concave parts, then the convex parts. The spraying movement can adopt an "S" shape or spiral movement. The surface layer spraying should be carried out within 8 hours after the completion of the base layer construction, generally controlled within 3 to 4 hours. The thickness of the surface layer spraying is 1 to 2cm. Before the surface layer spraying, the slope should be sprayed with water to ensure the adhesion between the base layer and the surface layer. Close-range hydroseeding should be carried out to ensure the uniformity of grass seed dispersal. Hydroseeding should be carried out from top to bottom, with the width of a single patch controlled at 4 to 6m.
[0030] Understandably, the three-dimensional vegetation net scheme in this embodiment utilizes living plants combined with engineering materials such as geosynthetics to construct a protective system with self-growth capabilities on the slope. This vegetation restoration technology reinforces the slope through plant growth. It employs a three-dimensional net made of biodegradable geosynthetic materials to fix the bare rock, followed by the spraying of a mixture of seeds, fertilizer, water-retaining agents, and clay. In the early stages of sowing, this mixture helps prevent erosion and retains soil to facilitate seed germination and growth. This scheme ensures that grass seeds are evenly distributed on the slope, preventing them from being lost due to wind and rain. It offers advantages such as rapid effectiveness, fewer seasonal limitations during construction, good soil stabilization performance, and good slope vegetation moisture retention.
[0031] Specifically, the process of the three-dimensional vegetation net solution includes slope treatment, laying the three-dimensional net, hydroseeding, covering, and maintenance. First, the slope is cleared of debris and stones, and low-lying areas are backfilled and compacted to ensure a smooth slope surface. The three-dimensional vegetation net is then laid along the slope from top to bottom, ensuring a smooth connection between the net and the slope surface. When laying the net at the top of the slope, it should extend 40-80cm, be buried in the soil, and compacted. The three-dimensional net is then fixed from bottom to top using U-shaped steel bars, spaced approximately 1.5-2.5m apart, with additional fixation using U-shaped iron nails or bamboo nails in between. After the three-dimensional vegetation net is laid, soil is evenly covered over it, ensuring no gaps appear and a soil thickness of at least 12mm. Then, fertilizer, growth hormone, and binder are mixed evenly in a specific ratio and applied to the surface. After backfilling with soil, hydraulic spraying is carried out. This involves mixing grass seeds (approximately 25 grams per square meter) with a growth promoter, wood fiber, fertilizer, growth hormone, moisture-retaining agent, and water in a specific ratio to form a homogeneous mixture. This mixture is then evenly sprayed onto the slope using a hydraulic sprayer. After spraying, a non-woven fabric is placed over the slope surface to retain moisture and reduce seed erosion from rainfall, promoting seed growth. If the temperature is too high, covering is unnecessary to prevent pests and diseases. Regular maintenance is necessary after spraying until the lawn is established. The non-woven fabric can be removed when the lawn reaches approximately 5 cm in height.
[0032] Understandably, the hole planting scheme in this embodiment is based on ecological geology theory, which involves artificially creating the necessary environment for plant growth by drilling planting holes in the rock face. The idea is to investigate and analyze the environmental conditions required for plant growth, and then modify the habitat conditions of steep rock slopes through scientific and reasonable engineering methods. This creates the necessary environmental conditions for plant growth on the steep rock slopes, and suitable plants are then planted on the modified steep rock face, allowing them to survive and take root. The plants then transition from initial artificial supplementation to drawing water and nutrients from the surrounding environment, ultimately returning to nature and achieving long-term greening of the steep rock slopes.
[0033] Specifically, the planting hole scheme involves three main processes: environmental space reconstruction, plant and soil reconstruction, plant planting, and maintenance management. First, environmental space reconstruction is performed to provide space for root development, primarily through the excavation of planting holes. This involves the angle, depth, and diameter of the planting holes. The angle is mainly related to terrain features, rock fissure development, and slope, requiring the holes to maximize the interception of rainfall and water-borne debris while excavating as many rock fissures as possible for ease of construction. The depth of the planting holes, considering plant habits, root development characteristics, rock fissure development, and construction conditions, can be determined to be around 50cm. The diameter of the planting holes needs to consider the space for root development and the amount of soil to be filled. If the diameter is too small, it obviously cannot provide enough space for root development, and too little soil will not provide the nutrient support necessary for the plant's self-organizing capabilities. However, if the diameter is too large, it increases construction difficulty and cost, and may also cause rock fragmentation. Taking all the above factors into consideration, a aperture of 20cm is appropriate. Plant soil regeneration involves several steps: First, the soil substrate. Ideally, the selected soil should be mature soil, meaning it has a history of planting or is cultivated soil. If cultivated soil is not feasible, the soil should be briefly soaked in water and mixed with an appropriate amount of sand to increase aeration. Second, soil fertility. Even cultivated soil has limited and incomplete fertility. Therefore, it is necessary to add appropriate fertilizers, such as nitrogen, phosphorus, and potassium, according to the habits of the selected species. Adding farmyard manure is best if conditions permit. It should be noted that more fertilizer is not necessarily better, as seedlings have limited fertility requirements. Third, soil moisture. Undoubtedly, soil moisture content is a key factor determining plant survival or natural growth. The moisture in the environment needs to be replenished appropriately during environment regeneration, and it also comes from within the rock mass. Therefore, the soil moisture content needs to be increased when filling the planting holes, but the amount of increase should be determined based on the moisture content of the local vegetation environment structure as investigated earlier. Planting involves the following steps: (1) Verify the permeability of the planting hole; holes with poor permeability are not suitable for planting; (2) Fill the hole with soil to 85% of its depth and compact it; (3) Plant the seedling in the hole, compact it first, and then lift the plant slightly to prevent its roots from being trapped; (4) Inject an appropriate amount of water into the hole. Final maintenance and management, in terms of the regeneration model, depends on the local climate. In arid and semi-arid areas, appropriate maintenance should be provided based on the plant's condition and rainfall. This means watering once a week after planting, and then deciding whether to water again based on the situation. However, once the plant has survived, watering is no longer necessary. In areas with abundant rainfall, such as south of the Yangtze River, theoretically, maintenance is not required unless extreme drought occurs.
[0034] It can be understood that the fish-scale pit solution in this embodiment is a revegetation method for slopes with many fissures in the slope rock mass and poor slope flatness. By using the concave parts of the slope or the spaces with small platforms, concrete or stones are stacked upwards or dug downwards to form pits, and planting pits are drilled or built at a certain density. Topsoil is filled in the pits to plant arbors, lianas or economic forest trees. This solution can effectively prevent rainwater scouring, conserve water and soil, with less construction earthwork, can be constructed all year round, is simple and easy to implement, has the characteristics of water conservation and drought resistance, and can combine arbors, shrubs, lianas and grasses.
[0035] Specifically, the fish-scale pit solution mainly includes slope cleaning, fish-scale pit masonry, backfilling of planting soil, vegetation planting and maintenance management. First, the stone dam foundation is cleared down to the bedrock. Then, the pit cavities approximately in the shape of a half moon are built with rubble and cement mortar, with the outer part higher and the inner part lower, and the long diameter is extended along the contour line direction. The top of the stone dam is plastered with fine aggregate concrete. During the construction process, the fish-scale pits are arranged in a "pin" shape, and the number of pits set per mu is determined according to the tree species to be planted. The backfilled planting soil is a vegetation growing substrate prepared by mixing soil, organic matter, compound fertilizer, water-retaining agent, biological fertilizer, etc., and is backfilled into the fish-scale pits. It is prohibited to dig mountain soil nearby to form fish-scale pit planting ponds. The soil thickness in the fish-scale pits should ensure sufficient growth of vegetation. Finally, arbors or lianas are planted in the fish-scale pit planting holes. In the later stage, it is maintained according to the greening specifications. In the early stage, attention should be paid to the water and fertilizer supply for the slope plants.
[0036] It can be understood that the floating platform greening solution in this embodiment refers to a technology in which holes are drilled in the rock slope, the anchor rods are anchored into the rock mass at an angle with the slope, and a reinforced concrete slab is poured on the upper part to make it in a "U" or "V" shape with the rock slope. Then, topsoil is covered inside to plant shrubs, vines, etc., so as to achieve the greening effect on the slope. This solution can protect the substrate from rainwater scouring, has good water and fertilizer retention effects, requires fine maintenance in the early stage, and is easier to maintain after the plants grow. The three-dimensional greening and overall greening effects are obvious.
[0037] Specifically, the greening scheme for the floating platform includes scaffolding preparation, platform construction, and planting. First, the rocky slope for the platform method is cleaned and reinforced using methods such as roughening, installing reinforcing mesh, and spraying concrete mixed with grass-green additives. Holes are drilled at pre-defined locations, anchoring reinforcing bars are installed, and mortar is injected into the holes. After the mortar reaches a certain strength, the reinforcing bars are welded together to form a steel skeleton. Then, formwork is hoisted and installed, the reinforcing bars are tied, and concrete is poured to form the platform structure, which is then cured to the design strength. A suitable growing substrate is prepared using soil, organic matter, compound fertilizer, water-retaining agent, and bio-fertilizer, and placed in the platform's planting troughs. Finally, climbing plants such as Virginia creeper are planted. A micro-sprinkler system is installed on the slope for watering. After planting, shade netting can be erected to facilitate seedling survival. The maintenance period should be no less than 45 days. During this period, timely watering, fertilization, and weeding are necessary. Sparse areas of seedlings should be replanted to avoid inconvenience to construction after the scaffolding is removed.
[0038] In this embodiment, the initial mapping includes a first mapping relationship, a second mapping relationship, a third mapping relationship, a fourth mapping relationship, a fifth mapping relationship, and a sixth mapping relationship.
[0039] Preferably, the first mapping relationship includes: when the restoration category is Class I restoration, the slope angle is 20°-50° and the slope type is soil slope or weakly weathered rock slope, the corresponding slope ecological restoration scheme is the plant soil nailing scheme.
[0040] Preferably, the second mapping relationship includes: when the repair category is Class I repair, the slope angle is greater than 60° and the slope type is weakly weathered rock slope or hard rock slope, the corresponding slope ecological restoration scheme is vegetation concrete scheme.
[0041] Preferably, the third mapping relationship includes: when the restoration category is Class I restoration, the slope angle is 30°-70° and the slope type is soil slope, weakly weathered rock slope or hard rock slope, the slope ecological restoration scheme is a three-dimensional vegetation network scheme.
[0042] Preferably, the fourth mapping relationship includes: when the restoration category is Class II restoration, the slope angle is greater than 70° and the slope type is hard rock slope, the slope ecological restoration scheme is the hole planting scheme.
[0043] Preferably, the fifth mapping relationship includes: when the restoration category is Class II restoration, the slope angle is 30°-55°, and the slope type is soil slope, weakly weathered rock slope or hard rock slope, the slope ecological restoration scheme is the fish scale pit scheme.
[0044] Preferably, the sixth mapping relationship includes: when the restoration category is Class II restoration, the slope angle is greater than 60°, and the slope type is a rock slope, the slope ecological restoration scheme is the floating platform greening scheme.
[0045] In this embodiment, when acquiring the slope angle and slope type of the slope to be restored, data can be collected through external acquisition devices, detection devices, or manual methods. For example, data can be collected through drone aerial surveys, geological drilling, and on-site reconnaissance. Simultaneously, the restoration category of the slope to be restored needs to be determined based on actual requirements. Other parameters, such as the slope area and shape, also need to be collected. Then, an initial solution is matched based on the collected data. The initial solution is based on a pre-set initial mapping. As described in the above embodiment, the initial mapping represents the correspondence between the restoration category, slope angle, and slope type and the ecological restoration solution for the slope. The initial mapping is pre-set based on historical data; that is, historical data is analyzed to determine the most suitable restoration solution for the slope under different conditions. Next, a slope matrix is generated based on the matched initial scheme. Specifically, an initial encoding matrix is generated based on the slope ecological restoration scheme, containing an encoding vector for each scheme. After obtaining the initial encoding matrix, the slope angle, slope type, and restoration category from the initial mapping are filled into it, resulting in the slope matrix. Finally, an encoder-decoder model is constructed. For example, the encoder has a 3-layer fully connected structure with 9-dimensional input and a 16-dimensional latent vector output, while the decoder has a 4-layer fully connected structure with 16-dimensional input and a 9-dimensional reconstruction matrix output. The encoder-decoder model is trained multiple times using several sets of historical data (historical slope restoration case data), for example, at least 100 times, to minimize the difference between the input matrix and the reconstruction matrix, such as the mean squared error. This allows the model to learn the deep correlation patterns between slope parameters and effective restoration schemes. The encoder is then removed from the encoder-decoder model, resulting in a trained decoder with 16-dimensional input. This decoder has the ability to generate a complete slope matrix from a 16-dimensional latent vector.
[0046] Step S2: Construct an evaluation model, determine the weights of evaluation indicators using the analytic hierarchy process, determine the evaluation scores of the slope ecological restoration scheme based on the evaluation indicator weights, and optimize the evaluation model according to the slope matrix and evaluation scores to obtain the optimized evaluation model.
[0047] The evaluation model is configured as a multilayer perceptron model, and the model structure corresponding to the evaluation model includes an input layer, an output layer, a first hidden layer, and a second hidden layer. The input layer has the same dimension as the slope matrix; the first hidden layer has 64 neurons and the second hidden layer has 32 neurons, both of which use the ReLU activation function; the output layer has 1 neuron to output the evaluation score, and the output layer uses the Sigmoid activation function.
[0048] Specifically, step S2 includes: Step S21: Construct an evaluation model based on a multilayer perceptron; Step S22: Determine the evaluation indicators for the slope to be repaired. The evaluation indicators include primary indicators and at least one secondary indicator corresponding to each primary indicator. Step S23: Based on the evaluation indicators, analyze the historical data using the analytic hierarchy process (AHP) to determine the weight vectors corresponding to each evaluation indicator, and calculate the evaluation score of the slope ecological restoration scheme using the weighted summation method based on the weight vectors. Step S24: Construct an optimization dataset using the slope matrix as the input sample and the evaluation score as the label; Step S25: Based on the optimized dataset, optimize the evaluation model using the mean squared error loss function and the Adam optimizer to obtain the optimized evaluation model.
[0049] First, the input layer of the evaluation model based on the multilayer perceptron is set to include 72 neurons to ensure consistency with the number of elements in the 9×8 dimensional slope matrix. The first hidden layer is set to include 64 neurons with the corresponding ReLU activation function, and the second hidden layer is set to include 32 neurons with the corresponding ReLU activation function. The output layer is set to include 1 neuron with the corresponding Sigmoid activation function. The output layer is used to output evaluation scores in the range [0,1]. The higher the score, the better the expected restoration effect of the corresponding slope ecological restoration scheme. After constructing the evaluation model based on the multilayer perceptron, the weights of the indicators are further determined. In this embodiment, the Analytic Hierarchy Process (AHP) is used to determine the weights of the indicators. The primary indicators include ecological benefits, engineering performance, economic costs, and construction feasibility. The secondary indicators corresponding to ecological benefits include vegetation restoration rate and soil erosion control rate. The secondary indicators corresponding to engineering performance include slope stability and erosion resistance. The secondary indicators corresponding to economic costs include material costs and labor costs. The secondary indicators corresponding to construction feasibility include construction difficulty and schedule adaptability. Experts from different fields can evaluate each primary indicator and its corresponding secondary indicators to obtain the indicator values for each evaluation indicator. Then, each primary indicator is compared to determine its weight vector. Further comparisons are made between the secondary indicators corresponding to any primary indicator to determine their respective weight vectors. After obtaining the weight vectors for different evaluation indicators, a weighted summation method is used to calculate the evaluation score for the corresponding slope ecological restoration scheme based on these weight vectors. Finally, using the corresponding slope matrix from historical data (historical slope restoration case data) as input and the calculated evaluation score as the target output, the evaluation model based on a multilayer perceptron is trained. The training objective is to minimize the error between the model's predicted score and the evaluation score calculated based on the weight vectors. After training, an optimized evaluation model is obtained, which can quickly evaluate any slope matrix.
[0050] In this embodiment, by establishing a structured multi-level evaluation index system and using the analytic hierarchy process (AHP) to quantify expert experience into scientific weights, a reliable scoring basis can be provided for the evaluation of slope ecological restoration schemes. The generated scores are used as labels to train a multilayer perceptron-based neural network model with the slope matrix as input. This model can autonomously learn the complex nonlinear relationships between various evaluation indicators and the final restoration effect from historical slope restoration case data. This not only significantly improves the automation and objectivity of slope ecological restoration scheme selection, but the trained evaluation model can also provide a reliable objective function, enabling intelligent search and iterative optimization of slope ecological restoration schemes. This embodiment, by deeply integrating the AHP and the multilayer perceptron model, constructs an intelligent scheme evaluation and optimization model, achieving a fundamental shift from subjective experience-driven to objective data-driven approaches, thereby improving the slope restoration effect.
[0051] Step S3: Generate the initial slope matrix iteratively through the decoder, evaluate the initial slope matrix through the optimized evaluation model, and obtain the optimal slope matrix based on the evaluation results.
[0052] Specifically, step S3 includes: Step S31: Construct the initial population based on the latent space dimension corresponding to the trained decoder; Step S32: Set the objective function with the goal of maximizing the evaluation score output by the evaluation model, and set the fitness function to map to the output score of the evaluation model; Step S33: Input each latent vector in the initial population into the trained decoder to obtain the initial slope matrix. In each iteration, multiple initial slope matrices are generated to form a matrix set. Step S34: Input the matrix set into the optimized evaluation model, output the evaluation score corresponding to each initial slope matrix through the first hidden layer and the second hidden layer, and sort them in descending order. Keep the initial slope matrix and its corresponding hidden vector with a score greater than the preset value as high-quality individuals. Step S35: Retain a preset number of high-quality individuals, and randomly mutate the latent vectors corresponding to the remaining high-quality individuals to form a new population; Step S36: Using the new population as input, iteratively execute steps S33 to S35. When the fluctuation of the highest evaluation score corresponding to consecutive m rounds is less than or equal to the difference threshold or the number of iterations reaches n rounds, terminate the iteration and select the initial slope matrix with the highest evaluation score as the optimal slope matrix.
[0053] Specifically, initialization is performed first. In the 16-dimensional latent space corresponding to the decoder, an initial population P0 containing 100 latent vectors is randomly generated. The objective function is set to maximize the score output by the evaluation model, and the fitness function is directly mapped to the evaluation model's score for the slope matrix generated by the decoder. Then, a genetic algorithm is used for iterative optimization. Each latent vector in the initial population P0 is input into the trained decoder to generate 100 initial slope matrices. All initial slope matrices are input into the optimized evaluation model to obtain their respective evaluation scores. They are sorted from high to low scores, and the top 30 initial slope matrices with high scores and their corresponding latent vectors are retained as high-quality individuals. The top 5 best individuals are directly entered into the next generation. A small random perturbation (i.e., random mutation) is applied to the latent vectors of the remaining 25 high-quality individuals to generate 95 new latent vectors, which are combined to form a new population P1 containing 100 individuals. Repeat the above decoding, evaluation, selection, and mutation process. When the highest score fluctuation is less than 0.001 for 10 consecutive iterations, or when the total number of iterations reaches 200, terminate the optimization and select the slope matrix with the highest evaluation score throughout the process as the optimal slope matrix.
[0054] In this embodiment, the population is randomly initialized in the low-dimensional latent space of the decoder. The decoder then transforms each latent vector into a complete slope matrix, and an evaluation model is used to score them. Subsequently, the optimal individual is selected and retained based on the score, and the latent vectors of the selected individuals are randomly mutated to obtain new situations, and this process is iterated repeatedly. This transforms the complex, multi-constrained slope repair scheme selection problem into an optimization problem with the goal of maximizing the score in a continuous latent space. First, it overcomes the limitations of simply relying on preset mapping rules, enabling intelligent exploration and innovation within a broad space of reasonable schemes defined by historical data distribution, further discovering better solutions. Second, the entire process is highly intelligent, greatly improving the efficiency and scale of scheme optimization, and handling a large number of parameter combinations. Finally, through elite retention and a score-based screening mechanism, it can stably converge to an optimal slope matrix with excellent overall performance, providing a reliable and quantitative decision-making basis for the final determination of the construction scheme, improving the slope repair effect and reducing repair costs.
[0055] Step S4: Decode the optimal slope matrix to determine the optimal ecological restoration scheme for the slope to be restored, and restore the slope to be restored according to the optimal ecological restoration scheme.
[0056] Further, step S4 includes: Step S41: Analyze the optimal slope matrix and extract the correspondence between the optimized restoration category, slope angle, and slope type and the ecological restoration scheme in the slope ecological restoration scheme; Step S42: Based on the optimized repair category, slope angle, and correspondence between slope type and slope ecological restoration scheme, rematch the slope ecological restoration scheme corresponding to the slope to be repaired as the optimal slope ecological restoration scheme. Step S43: Set the final restoration plan for the slope to be restored based on the optimal ecological restoration plan, so as to achieve ecological restoration of the slope to be restored.
[0057] Specifically, the optimal slope matrix is first analyzed to extract the scheme rows of the encoded vectors in the optimal slope matrix. For example, decoding confirms that the optimized scheme type is still the hole planting scheme, but some related parameters have changed. For example, under the condition that the slope type is hard rock slope and the repair type is Class II repair, the slope angle obtained after analysis has changed from 70° to 75°. If the current slope angle of the slope to be repaired is 72°, it means that the slope to be repaired is no longer suitable for the hole planting scheme. In this case, it is necessary to re-analyze and evaluate each slope ecological repair scheme based on historical data (historical slope repair case data), redetermine the parameters corresponding to each slope ecological repair scheme to optimize the mapping relationship between each slope ecological repair scheme and slope parameters, match the slope ecological repair scheme corresponding to the slope to be repaired according to the optimized mapping relationship, and take it as the optimal slope ecological repair scheme for the slope to be repaired. Set the final repair scheme for the slope to be repaired according to the optimal slope ecological repair scheme, and execute the final repair scheme to carry out ecological repair on the slope to be repaired.
[0058] In this embodiment, by parsing the optimal slope matrix, specific slope ecological restoration schemes and corresponding slope parameters can be extracted. Based on the mapping relationship between the decoded optimized slope parameters and the slope ecological restoration schemes, when the mapping relationship does not match the actual slope parameters, a dynamic update mechanism based on historical data is used to re-evaluate and optimize the correspondence rules between all slope ecological restoration schemes and slope parameters. This allows for the matching of the truly applicable optimal scheme for the current slope to be restored, improving the reliability of the original experience knowledge base built based on historical data. Ultimately, a highly reliable restoration scheme is generated, significantly enhancing the reliability of the slope ecological method in this embodiment and improving the effect of slope ecological restoration.
[0059] like Figure 2 As shown, according to another aspect of this application, a slope ecological restoration device is also provided, which is used to perform the above-described slope ecological restoration method. The device includes: Decoder training module 10 is used to determine the ecological restoration scheme for the slope to be restored, generate a slope matrix based on the ecological restoration scheme, and train a deep learning model based on the slope matrix to obtain a trained decoder. The model generation and optimization module 20 is used to construct the evaluation model. The weights of the evaluation indicators are determined by the analytic hierarchy process. The evaluation scores of the slope ecological restoration scheme are determined based on the weights of the evaluation indicators. The evaluation model is optimized according to the slope matrix and the evaluation scores to obtain the optimized evaluation model. The optimal matrix filtering module 30 is used to iteratively generate an initial slope matrix through a decoder, evaluate the initial slope matrix through an optimized evaluation model, and obtain the optimal slope matrix based on the evaluation results. The repair scheme determination module 40 is used to decode the optimal slope matrix, determine the optimal ecological restoration scheme for the slope to be restored, and restore the slope to be restored according to the optimal ecological restoration scheme.
[0060] like Figure 3 As shown, according to another aspect of this application, an electronic device 100 is also provided, which includes: a processor 101 and a memory 103 storing a program, the program including instructions that, when executed by the processor 101, cause the processor 101 to perform the above-described slope ecological restoration method.
[0061] According to another aspect of this application, a non-transitory machine-readable medium storing computer instructions for causing the computer to perform the above-described slope ecological restoration method is also provided.
[0062] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.
[0063] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore should not be construed as limiting the scope of protection of this invention.
[0064] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for ecological restoration of slopes, characterized in that, include: Step S1: Determine the ecological restoration scheme for the slope to be restored, generate a slope matrix based on the ecological restoration scheme, train a deep learning model based on the slope matrix, and obtain a trained decoder. Step S2: Construct an evaluation model, determine the weights of evaluation indicators using the analytic hierarchy process, determine the evaluation score of the slope ecological restoration scheme based on the evaluation indicator weights, and optimize the evaluation model according to the slope matrix and the evaluation score to obtain the optimized evaluation model; Step S3: The initial slope matrix is generated iteratively through the decoder, and the initial slope matrix is evaluated through the optimized evaluation model. Based on the evaluation results, the optimal slope matrix is obtained. Step S4: Decode the optimal slope matrix to determine the optimal ecological restoration scheme for the slope to be restored, and restore the slope to be restored according to the optimal ecological restoration scheme.
2. The slope ecological restoration method according to claim 1, characterized in that, Step S1 includes: Step S11: Obtain the repair category, slope angle, and slope type corresponding to the slope to be repaired; Step S12: Based on the repair category, the slope angle, and the slope type, obtain the slope ecological restoration plan corresponding to the slope to be repaired through a preset initial mapping relationship; The initial mapping relationship is configured as the correspondence between the restoration category, the slope angle, the slope type, and the slope ecological restoration scheme; Step S13: Construct an initial coding matrix based on the slope ecological restoration scheme. Fill in the initial coding matrix to obtain the slope matrix according to the correspondence between the slope angle, the slope type, the restoration category and the slope ecological restoration scheme. Step S14: Construct a deep learning model based on an encoder-decoder structure, train the deep learning model based on historical data, and remove the decoder module from the trained deep learning model to obtain the trained decoder.
3. The slope ecological restoration method according to claim 2, characterized in that, The restoration categories include Class I restoration and Class II restoration. Class I restoration is configured to improve the stability of the slope to be restored, and Class II restoration is configured to create habitats for the slope to be restored; and / or, The slope types include at least soil slopes, weakly weathered rock slopes, and hard rock slopes; and / or, The slope ecological restoration plan includes at least the following: plant nailing plan, vegetation concrete plan, three-dimensional vegetation net plan, hole planting plan, fish scale pit plan, and ledge greening plan.
4. The slope ecological restoration method according to claim 2, characterized in that, The evaluation model is configured as a multilayer perceptron model, and the model structure corresponding to the evaluation model includes an input layer, an output layer, a first hidden layer, and a second hidden layer. The input layer is configured with the same dimension as the slope matrix; the first hidden layer includes 64 neurons, the second hidden layer includes 32 neurons, and the activation functions of both the first and second hidden layers are ReLU functions; the output layer includes 1 neuron for outputting the evaluation score, and the activation function of the output layer is the Sigmoid function.
5. The slope ecological restoration method according to claim 4, characterized in that, Step S2 includes: Step S21: Construct an evaluation model based on a multilayer perceptron; Step S22: Determine the evaluation index of the slope to be repaired, wherein the evaluation index includes primary indexes and at least one secondary index corresponding to each primary index; Step S23: Based on the evaluation indicators, analyze the historical data using the analytic hierarchy process (AHP) to determine the weight vector corresponding to each evaluation indicator, and calculate the evaluation score of the slope ecological restoration scheme based on the weight vector using the weighted summation method. Step S24: Use the slope matrix as the input sample and the evaluation score as the label to construct an optimization dataset; Step S25: Based on the optimized dataset, optimize the evaluation model using the mean squared error loss function and the Adam optimizer to obtain the optimized evaluation model.
6. The slope ecological restoration method according to claim 5, characterized in that, Step S3 includes: Step S31: Construct an initial population based on the latent space dimension corresponding to the trained decoder; Step S32: Set an objective function with the goal of maximizing the evaluation score output by the evaluation model, and set the fitness function to map to the output score of the evaluation model; Step S33: Input each latent vector in the initial population into the trained decoder to obtain the initial slope matrix. In each iteration, generate multiple initial slope matrices to form a matrix set. Step S34: Input the set of matrices into the optimized evaluation model, output the evaluation score corresponding to each initial slope matrix through the first hidden layer and the second hidden layer, and sort them in descending order. The initial slope matrices with scores greater than the preset value and their corresponding hidden vectors are retained as high-quality individuals. Step S35: Retain a preset number of high-quality individuals, and randomly mutate the latent vectors corresponding to the remaining high-quality individuals to form a new population; Step S36: Using the new population as input, iteratively execute steps S33 to S35. When the fluctuation of the highest evaluation score corresponding to m consecutive rounds is less than or equal to the difference threshold or the number of iterations reaches n rounds, terminate the iteration and select the initial slope matrix with the highest evaluation score as the optimal slope matrix.
7. The slope ecological restoration method according to claim 2, characterized in that, Step S4 includes: Step S41: Analyze the optimal slope matrix and extract the correspondence between the optimized restoration category, the slope angle, and the slope type and the ecological restoration scheme in the slope ecological restoration scheme; Step S42: Based on the optimized correspondence between the repair category, the slope angle, the slope type, and the slope ecological restoration scheme, rematch the slope ecological restoration scheme corresponding to the slope to be restored as the optimal slope ecological restoration scheme; Step S43: Set the final restoration plan for the slope to be restored according to the optimal slope ecological restoration plan, so as to realize the ecological restoration of the slope to be restored.
8. A slope ecological restoration device, used to perform the slope ecological restoration method according to any one of claims 1 to 7, characterized in that, include: The decoder training module is used to determine the ecological restoration scheme corresponding to the slope to be restored, generate a slope matrix based on the ecological restoration scheme, and train a deep learning model based on the slope matrix to obtain a trained decoder. The model generation and optimization module is used to construct an evaluation model, determine the weights of evaluation indicators through the analytic hierarchy process, determine the evaluation score of the slope ecological restoration scheme based on the evaluation indicator weights, and optimize the evaluation model according to the slope matrix and the evaluation score to obtain the optimized evaluation model. An optimal matrix filtering module is used to iteratively generate an initial slope matrix through the decoder, evaluate the initial slope matrix through an optimized evaluation model, and obtain the optimal slope matrix based on the evaluation results. The repair scheme determination module is used to decode the optimal slope matrix, determine the optimal ecological restoration scheme for the slope to be restored, and restore the slope to be restored according to the optimal ecological restoration scheme.
9. An electronic device, comprising: A processor and a memory storing a program, characterized in that the program includes instructions that, when executed by the processor, cause the processor to perform the slope ecological restoration method according to any one of claims 1 to 7.
10. A non-transitory machine-readable medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the slope ecological restoration method according to any one of claims 1 to 7.