Land ecological restoration method, system, equipment and medium

Abstract

The invention provides a land ecological restoration method, system and equipment and a medium. The method comprises the following steps: acquiring multi-source monitoring data of a set area; based on the multi-source monitoring data, performing land utilization prediction on a set area by using the dynamic prediction model to obtain a land utilization change prediction result; and if a land utilization change prediction result exceeds a land development early warning range, taking ecological environment limitation as a constraint condition, taking an optimal ecological restoration result as an objective function, and solving the objective function by using a multi-objective genetic algorithm to obtain an ecological restoration scheme of the set region. The method can quickly determine whether the land is over-developed or not, improves the efficiency and accuracy of land supervision, can generate an optimal ecological restoration scheme when the land is over-developed, can improve the restoration efficiency, reduces the restoration cost, can provide a sustainable solution for regional ecological protection, and improves the economic benefit. And the robustness and adaptability of the system are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent management of land resources, and particularly relates to a land ecological restoration method, system, device and medium. Background Art

[0002] Land comprehensive improvement refers to optimizing the land use structure through engineering measures, while ecological restoration focuses on restoring damaged ecological functions. In the traditional land comprehensive improvement and ecological restoration process, remote sensing monitoring technology is first used to obtain land cover data, then the effect of land improvement is evaluated and predicted through spatial analysis, and finally on-site manual verification is carried out to accept the project quality of land improvement and ecological restoration.

[0003] However, the following problems exist in the traditional land comprehensive improvement and ecological restoration process:

[0004] 1. The update cycle of satellite remote sensing data is long, for example, generally 5 - 7 days. The too long data update cycle makes it difficult to capture dynamic changes;

[0005] 2. In traditional treatment schemes, static models are mostly used to predict the effect of land improvement, and static models cannot accurately predict the treatment effect;

[0006] 3. There are manual intervention nodes in the supervision process. The average project approval cycle in the manual access nodes exceeds 30 days, which will lead to loss of supervision efficiency.

[0007] In summary, traditional land improvement and ecological restoration have the disadvantages of low accuracy and poor efficiency. Summary of the Invention

[0008] In order to overcome the defects of low accuracy and poor efficiency in the above-mentioned traditional land improvement and ecological restoration, the present invention provides a land ecological restoration method, including:

[0009] Obtaining multi-source monitoring data of a set area;

[0010] Based on the multi-source monitoring data, using a dynamic prediction model to perform land use prediction on the set area to obtain a land use change prediction result;

[0011] If the land use change prediction result exceeds the land development warning range, taking the ecological environment restriction as a constraint condition and the optimal ecological restoration result as an objective function, using a multi-objective genetic algorithm to solve the objective function to obtain an ecological restoration plan for the set area.

[0012] Optionally, the step of based on the multi-source monitoring data, using a dynamic prediction model to perform land use prediction on the set area to obtain a land use change prediction result includes:

[0013] Based on the multi-source monitoring data, a multi-modal feature fusion network is used for land use classification to obtain the land types of the set area;

[0014] Based on the multi-source monitoring data and the land types, a dynamic prediction sub-module is used for land use change prediction to obtain a land use change prediction result.

[0015] Optionally, the using the multi-modal feature fusion network in the dynamic prediction model for land use classification based on the multi-source monitoring data to obtain the land types of the set area includes:

[0016] Using the three-dimensional convolutional layer of the multi-modal feature fusion network in the dynamic prediction model to extract features from the hyperspectral image and lidar point cloud data in the multi-source monitoring data to obtain a spatial-spectral joint feature; the spatial-spectral joint feature includes a spectral reflectance feature and a terrain elevation feature;

[0017] Using the cross-modal attention gating mechanism of the multi-modal feature fusion network to assign weights to the spectral reflectance feature and the terrain elevation feature to obtain a spectral feature weight and a terrain feature weight;

[0018] Based on the spectral reflectance feature, the terrain elevation feature, the spectral feature weight, and the terrain feature weight, the output layer of the multi-modal feature fusion network is used for land use classification to obtain the land types of the set area.

[0019] Optionally, the using the cross-modal attention gating mechanism of the multi-modal feature fusion network to assign weights to the spectral reflectance feature and the terrain elevation feature to obtain a spectral feature weight and a terrain feature weight includes:

[0020] Using the cross-modal attention gating mechanism of the multi-modal feature fusion network and combining with a dynamic weight assignment function to assign weights to the spectral reflectance feature and the terrain elevation feature to obtain a spectral feature weight and a terrain feature weight;

[0021] The dynamic weight assignment function satisfies the following formula:

[0022]

[0023] Q2 = 1 - Q1;

[0024] Wherein, represents vector concatenation, S ′ represents the spectral reflectance feature, T ′ represents the terrain elevation feature, W g is a gating weight matrix, δ is the Sigmoid function, Q1 is the spectral feature weight, and Q2 is the terrain feature weight.

[0025] Optionally, the multi-objective genetic algorithm is a parallel multi-objective genetic algorithm.

[0026] Optionally, taking the ecological environment restriction as a constraint condition and the optimal ecological restoration result as an objective function, using the multi-objective genetic algorithm to solve the objective function, the ecological restoration plan for the set area includes:

[0027] Taking the ecological environment restriction as a constraint condition, initializing the population individuals of the multi-objective genetic algorithm; the initialized population individuals include the initial ecological restoration plan;

[0028] Calculating the fitness values of the initialized population individuals according to the fitness function, performing multiple crossovers and mutations on the initialized population individuals to obtain new population individuals; iteratively updating the population individuals until the maximum iteration termination condition is reached, and taking the ecological restoration plan corresponding to the new population individuals at this time as the ecological restoration plan corresponding to the optimal ecological restoration result of the objective function;

[0029] The constraint conditions include vegetation coverage rate constraint, soil erosion constraint, biodiversity constraint, soil heavy metal content constraint value and surface runoff constraint.

[0030] Optionally, after taking the ecological environment restriction as a constraint condition and the optimal ecological restoration result as an objective function, using the multi-objective genetic algorithm to solve the objective function to obtain the ecological restoration plan for the set area, it further includes:

[0031] Based on the ecological restoration plan, carrying out ecological restoration on the set area;

[0032] Based on the multi-source monitoring data of the set area after ecological restoration, using the Bayesian network evaluation model to evaluate the restoration degree to obtain the ecological restoration degree;

[0033] Based on the ecological restoration degree, carrying out ecological restoration supervision on the set area.

[0034] On the other hand, the present invention also provides a land ecological restoration system, including:

[0035] An acquisition module, configured to acquire multi-source monitoring data of a set area;

[0036] A land use prediction module, configured to perform land use prediction on the set area based on the multi-source monitoring data by using a dynamic prediction model to obtain a land use change prediction result;

[0037] An ecological restoration module, which is used to, if the land use change prediction result exceeds the land development warning range, take the ecological environment constraint as a constraint condition and the optimal ecological restoration result as an objective function, and use a multi-objective genetic algorithm to solve the objective function to obtain the ecological restoration plan for the set area.

[0038] Optionally, the land use prediction module is specifically configured to classify the land use based on the multi-source monitoring data by using a multi-modal feature fusion network to obtain the land types of the set area; and predict the land use change based on the multi-source monitoring data and the land types by using a dynamic prediction sub-module to obtain the land use change prediction result.

[0039] Optionally, the land use prediction module is specifically configured to use the three-dimensional convolutional layer of the multi-modal feature fusion network in the dynamic prediction model to extract features from the hyperspectral image and the lidar point cloud data in the multi-source monitoring data to obtain the spatial-spectral joint features; the spatial-spectral joint features include spectral reflectance features and terrain elevation features; use the cross-modal attention gating mechanism of the multi-modal feature fusion network to assign weights to the spectral reflectance features and the terrain elevation features to obtain spectral feature weights and terrain feature weights; and classify the land use based on the spectral reflectance features, the terrain elevation features, the spectral feature weights, and the terrain feature weights by using the output layer of the multi-modal feature fusion network to obtain the land types of the set area.

[0040] Optionally, the land use prediction module is specifically configured to use the cross-modal attention gating mechanism of the multi-modal feature fusion network and combine it with a dynamic weight assignment function to assign weights to the spectral reflectance features and the terrain elevation features to obtain spectral feature weights and terrain feature weights.

[0041] The dynamic weight assignment function satisfies the following formula:

[0042]

[0043] Q2 = 1 - Q1;

[0044] Where represents vector concatenation, S ′ represents the spectral reflectance feature, T ′ represents the terrain elevation feature, W g is the gating weight matrix, δ is the Sigmoid function, Q1 is the spectral feature weight, and Q2 is the terrain feature weight.

[0045] Optionally, the multi-objective genetic algorithm is a parallel multi-objective genetic algorithm.

[0046] Optionally, the ecological restoration module is specifically configured to initialize the population individuals of the multi-objective genetic algorithm with ecological environment constraints as constraints; the initialized population individuals include initial ecological restoration plans; calculate the fitness values of the initialized population individuals according to the fitness function, perform multiple crossovers and mutations on the initialized population individuals to obtain new population individuals; repeatedly iterate and update the population individuals until the maximum iteration termination condition is reached, and use the ecological restoration plan corresponding to the new population individuals at this time as the ecological restoration plan corresponding to the optimal ecological restoration result of the objective function; the constraints include vegetation coverage constraint, soil erosion constraint, biodiversity constraint, soil heavy metal content constraint value, and surface runoff constraint.

[0047] Optionally, the device further includes:

[0048] An ecological restoration supervision module, configured to perform ecological restoration on the set area based on the ecological restoration plan; evaluate the restoration degree using a Bayesian network evaluation model based on multi-source monitoring data of the set area after ecological restoration to obtain the ecological restoration degree; perform ecological restoration supervision on the set area based on the ecological restoration degree.

[0049] On the other hand, the present invention also provides an electronic device, including: at least one processor and a memory; the memory and the processor are connected by a bus;

[0050] The memory is used to store one or more programs;

[0051] When the one or more programs are executed by the at least one processor, the land ecological restoration method described in any one of the above is implemented.

[0052] On the other hand, the present invention also provides a readable storage medium, on which an execution program is stored, and when the execution program is executed, the land ecological restoration method described in any one of the above is implemented.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] The present invention provides a method, system, device and medium for land ecological restoration. The method includes: obtaining multi-source monitoring data of a set area; based on the multi-source monitoring data, using a dynamic prediction model to predict land use in the set area to obtain a prediction result of land use change; if the prediction result of land use change exceeds the land development warning range, using ecological environment constraints as constraint conditions and the optimal ecological restoration result as the objective function, and using a multi-objective genetic algorithm to solve the objective function to obtain an ecological restoration plan for the set area. The present invention can quickly determine whether there is overdevelopment of land, improve the efficiency and accuracy of land supervision, and can generate an optimal ecological restoration plan when there is overdevelopment, can also improve the restoration efficiency, reduce the restoration cost, can provide a sustainable solution for regional ecological protection, and enhance the robustness and adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a schematic flowchart of the land ecological restoration method of the present invention;

[0056] Figure 2 is a schematic structural diagram of the land ecological restoration system of the present invention;

[0057] Figure 3 is a schematic structural diagram of the electronic device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] The following further describes in detail the specific embodiments of the present invention with reference to the drawings.

[0059] Example 1:

[0060] A land ecological restoration method provided by the present invention, the schematic flowchart is as Figure 1 shown, and includes:

[0061] Step 101: Obtain multi-source monitoring data of a set area;

[0062] Step 102: Based on the multi-source monitoring data, use a dynamic prediction model to predict land use in the set area to obtain a prediction result of land use change;

[0063] Step 103: If the prediction result of land use change exceeds the land development warning range, use ecological environment constraints as constraint conditions and the optimal ecological restoration result as the objective function, and use a multi-objective genetic algorithm to solve the objective function to obtain an ecological restoration plan for the set area.

[0064] A land ecological restoration method provided by an embodiment of the present invention is applied to an electronic device, and the electronic device can be a personal computer (PC), a server, etc.

[0065] In an embodiment of the present invention, a drone can carry a multispectral sensor and a lidar to collect data from a set area and obtain multi-source detection data of the set area. Among them, the multi-source detection data includes, but is not limited to, hyperspectral images collected by the multispectral sensor and lidar point cloud data collected by the lidar. Among them, the hyperspectral images can cover the visible light, near-infrared, and short-wave infrared bands, and the spatial resolution is 10 meters. To ensure data quality, after the multispectral sensor collects image data, atmospheric correction and geometric correction are performed on the image data to obtain hyperspectral images. The point cloud density of the lidar point cloud data can be 50 points per square meter. To ensure data quality, after the lidar collects point cloud data, denoising and classification processing can also be performed on the point cloud data to obtain lidar data.

[0066] Optionally, after the electronic device obtains the multi-source monitoring data, it can also preprocess the hyperspectral images and lidar point cloud data carried in the multi-source monitoring data to ensure data quality and provide reliable input for subsequent processing.

[0067] For example, denoising and normalization processing are performed on the hyperspectral images, and filtering and interpolation processing are performed on the lidar point cloud data.

[0068] In an implementation of the present invention, the electronic device inputs the multi-source monitoring data into a dynamic prediction model, so that the dynamic prediction model predicts the land use change situation of the set area corresponding to the multi-source monitoring data, and the electronic device obtains the land use change prediction result output by the dynamic prediction model.

[0069] For example, the land use change prediction result can be the change in land use rate, or the change in the area of each land type, etc.

[0070] The electronic device is also preconfigured with a land development warning range, which is used to determine whether there is overdevelopment of the land. If the land development result exceeds the land development warning range, an alarm is issued, and ecological restoration is required. If the land development result does not exceed the land development warning range, no processing is required.

[0071] Based on this, after the electronic device obtains the land change prediction result using the dynamic prediction model, it can determine whether the land change prediction result exceeds the land development warning range. If the land use change prediction result exceeds the land development warning range, an ecological restoration plan is generated.

[0072] Exemplarily, the electronic device uses ecological environment constraints as constraint conditions, takes the optimal ecological restoration result as the objective function, and uses a multi-objective genetic algorithm to solve the objective function, thereby obtaining the optimal solution corresponding to the objective function, and determining the ecological restoration plan for the set area as the solution corresponding to the optimal solution.

[0073] Among them, the ecological environment constraints are ecological environment parameters, including but not limited to vegetation coverage rate, soil erosion value, biodiversity, etc.

[0074] In the present invention, based on multi-source monitoring data of the set area, a land use change prediction result is predicted by using a dynamic prediction model, so that it is possible to determine whether there is overdevelopment according to the land use change prediction result and the preset land development warning range, and the monitoring of land development is realized quickly and efficiently. At the same time, if it is predicted that there is overdevelopment, an optimal ecological restoration plan will also be determined based on ecological environment constraints by using a multi-objective genetic algorithm, realizing the ecological restoration of the supervised area.

[0075] In order to realize the prediction of land use change, on the basis of the above-mentioned embodiment, in the embodiment of the present invention, the above-mentioned land use prediction of the set area by using a dynamic prediction model based on multi-source monitoring data to obtain a land use change prediction result includes:

[0076] Based on multi-source monitoring data, use the multi-modal feature fusion network of the dynamic prediction model to classify land use and obtain the land type of the set area;

[0077] Based on multi-source monitoring data and land type, use the dynamic prediction sub-module to predict land use change and obtain the land use change prediction result.

[0078] In the embodiment of the present invention, the dynamic prediction model includes a multi-modal feature fusion network and a dynamic prediction sub-module. Among them, the multi-modal feature fusion network is used for land classification, and the dynamic prediction sub-module is used for land use change prediction.

[0079] Exemplarily, the multi-modal feature fusion network is constructed using a deep learning framework (such as an improved U-Net network), and the model structure is a Convolutional Neural Networks (CNN). The input of the multi-modal feature fusion network is multi-source monitoring data. The multi-modal feature fusion network extracts and fuses features from the multi-source monitoring data, and based on the attention mechanism, classifies and identifies by combining the fused features, and outputs the land type of the set area. Among them, the land type includes but not limited to cultivated land, forest land, grassland, water area, construction land, etc.

[0080] In the present invention, the multi-modal feature fusion network is trained using historical annotation data, and the model parameters are optimized through cross-validation to ensure that the classification accuracy reaches over 95%. The historical annotation data is 1 million + annotation samples from all over the country, covering various terrains, enabling the land classification model to adapt to different terrain areas (plain / hill / mountain) through regional fine-tuning.

[0081] In the present invention, the dynamic prediction sub-module analyzes the change trend of historical land use data, combines regional development plans and natural environmental factors, and predicts the changes in land use in the next period of time.

[0082] For example, the dynamic prediction sub-module predicts that a set area will change from forest land to construction land within the next five years. The prediction result generates a heat map of land use changes, and different colors in the heat map represent the intensity and scope of land use changes. For example, red represents areas with drastic changes, and green represents areas with minor changes.

[0083] For example, the dynamic prediction sub-module adopts a Long Short-Term Memory (LSTM) model. The input of this dynamic prediction sub-module is multi-source monitoring data and land types. This dynamic prediction sub-module analyzes historical data and generates a heat map of land use changes within a preset time range, such as generating a heat map of land use changes within the next 3 years.

[0084] The electronic device can automatically determine whether the set area may exceed the ecological carrying capacity according to the preset land development warning range and the prediction result of land use changes, and issue a warning for the development red line.

[0085] For example, when the prediction result of land use changes in the set area exceeds the land development warning range, a warning signal will be issued to remind relevant departments to take measures to restrict development activities to protect the ecological environment.

[0086] Moreover, in the embodiments of the present invention, the warning signal is displayed through a visual interface to help decision-makers quickly identify problem areas and take corresponding measures.

[0087] For example, three-dimensional real-scene modeling is realized through the integration of the Cesium engine on a dynamic data visualization platform, supporting the overlay display of multi-layers of historical / real-time / prediction data; if the dynamic prediction sub-module determines that a level-four alarm is triggered according to the preset land development warning range (such as the vegetation degradation rate > 15%), customized decision-making suggestions will be pushed to the mobile terminal.

[0088] In order to achieve land classification and improve the efficiency of land classification, based on the above embodiments, in the embodiments of the present invention, the above-mentioned land use classification is carried out by using the multi-modal feature fusion network in the dynamic prediction model based on multi-source monitoring data, and the land types in the set area are obtained as follows:

[0089] Using the three-dimensional convolutional layer of the multi-modal feature fusion network in the dynamic prediction model, feature extraction is performed on the hyperspectral image and lidar point cloud data in the multi-source monitoring data to obtain spatial-spectral joint features; the spatial-spectral joint features include spectral reflectance features and terrain elevation features;

[0090] Using the cross-modal attention gating mechanism of the multi-modal feature fusion network, weights are assigned to the spectral reflectance feature and the terrain elevation feature to obtain spectral feature weights and terrain feature weights;

[0091] Based on the spectral reflectance feature, terrain elevation feature, spectral feature weights and terrain feature weights, land use classification is performed using the output layer of the multi-modal feature fusion network to obtain the land types in the set area.

[0092] In the present invention, the multi-modal feature fusion network uses an improved U-Net network for land use classification and integrates the attention mechanism to improve the accuracy of land classification. For example, the multi-modal feature fusion network may include a three-dimensional convolutional layer (Conv3D), a cross-modal attention gating mechanism, and an output layer.

[0093] In the present invention, the multi-source monitoring data includes but is not limited to hyperspectral images and lidar point clouds. Among them, the hyperspectral image provides rich spectral information and can accurately identify the spectral characteristics of different ground objects, while the lidar point cloud data provides high-precision terrain information and can reflect the subtle undulations of the ground surface.

[0094] Based on this, in the present invention, the three-dimensional convolutional layer of the multi-modal feature fusion network is used for feature extraction of multi-source monitoring data, the cross-modal attention gating mechanism is used to assign weights to each feature output by the three-dimensional convolutional layer, and the output layer is used for land use classification based on the outputs of the three-dimensional convolutional layer and the cross-modal attention gating mechanism.

[0095] For example, the multi-modal feature fusion network can first use the three-dimensional convolutional layer to perform feature extraction on the hyperspectral image and extract the spectral reflectance feature from the hyperspectral image. This spectral reflectance feature can reflect the spectral characteristics of different ground objects. For example, vegetation has a high reflectance in the near-infrared band and water has a low reflectance in the short-wave infrared band. Among them, the three-dimensional convolutional layer can capture the correlations in both the spatial neighborhood and spectral dimensions at the same time. For example: the first convolutional layer: uses a small convolutional kernel (such as 3x3x7) to extract local spatial-spectral features; subsequent convolutional layers: gradually increase the receptive field to extract more abstract features.

[0096] Meanwhile, the 3D convolutional layer processes the lidar point cloud data and extracts terrain elevation features from it. These terrain elevation features can reflect the undulation changes of the terrain, such as the steepness of slopes and the flat areas of valleys.

[0097] In the embodiments of the present invention, the land classification model uses a cross-modal attention gating mechanism to assign weights to the spectral reflectance features and terrain elevation features, obtaining spectral feature weights and terrain feature weights. And based on the spectral feature weights and terrain feature weights, it performs feature fusion on the spectral reflectance features and terrain elevation features to obtain fused features.

[0098] Exemplarily, the spectral feature weights and terrain feature weights of the spectral reflectance features and terrain elevation features are calculated through the cross-modal attention gating mechanism, and they are weighted and summed according to the weights to obtain fused features. Among them, the parameters of the cross-modal attention gating mechanism are learned from the training data and can dynamically adjust the corresponding weights according to the importance of the two features. For example, in the vegetation classification task, the calculated spectral feature weights may be higher; while in the terrain classification task, the calculated terrain feature weights may be higher.

[0099] In the present invention, the output layer of the multi-modal feature fusion network is used for land use classification. Exemplarily, this output layer can be a classifier, and this classifier can be a support vector machine or a random forest, etc. The classifier performs classification and recognition based on the fused features. This output layer learns the feature distributions of different land types from the training data and outputs a target map carrying the land types. This target map can clearly show the distribution of different land types within the region, such as the concentrated areas of cultivated land, the distribution ranges of forest land, and the specific locations of water areas, providing data support for subsequent ecological restoration.

[0100] Among them, the process of feature fusion can also be completed in the output layer.

[0101] In order to improve the accuracy of land classification, based on the above embodiments, in the embodiments of the present invention, the above-mentioned use of the cross-modal attention gating mechanism of the multi-modal feature fusion network to assign weights to the spectral reflectance features and terrain elevation features, obtaining spectral feature weights and terrain feature weights includes:

[0102] Using the cross-modal attention gating mechanism of the multi-modal feature fusion network, combined with a dynamic weight assignment function, to assign weights to the spectral reflectance features and terrain elevation features, obtaining spectral feature weights and terrain feature weights;

[0103] The dynamic weight assignment function satisfies the following formula:

[0104]

[0105] Q2 = 1 - Q1;

[0106] Wherein, denotes vector concatenation, S ′ denotes spectral reflectance feature, T ′ denotes terrain elevation feature, W g is the gating weight matrix, δ is the Sigmoid function, Q1 is the spectral feature weight, and Q2 is the terrain feature weight.

[0107] In the present invention, before using the cross-modal attention gating mechanism to assign weights to the spectral reflectance feature and the terrain elevation feature, the spectral reflectance feature and the terrain elevation feature can be respectively subjected to feature projection by using a preset projection matrix. The projection matrix is learned from training data and can map high-dimensional features to a low-dimensional space, thereby reducing the computational complexity.

[0108] By using the cross-modal attention gating mechanism and combining with the dynamic weight assignment function, weights are assigned to the spectral reflectance feature and the terrain elevation feature to obtain the spectral feature weight and the terrain feature weight, and the spectral reflectance feature and the terrain elevation feature are weighted and summed according to the spectral feature weight and the terrain feature weight to obtain the fused feature. The fused feature can simultaneously reflect the spectral characteristics and terrain features of the ground object, thereby improving the accuracy of land classification.

[0109] Exemplarily, the electronic device can obtain the spectral feature weight and the terrain feature weight through the following formula:

[0110] 1. Feature projection:

[0111] Let the spectral reflectance feature be S ∈ R d×1 , and the terrain elevation feature be T ∈ R d×1 , that is, let the spectral reflectance feature and the terrain elevation feature be matrices with d rows and 1 column, and they are mapped to the shared space through a learnable matrix:

[0112] S ′ = W1 · S + b1;

[0113] T ′ = W2 · T + b2;

[0114] Wherein, W1, W2 ∈ R d×d , W1 is the first projection matrix, W2 is the second projection matrix, that is, W1 and W2 are matrices with d rows and d columns, b1 and b2 are bias terms, S ′ is the projected spectral reflectance feature, and T ′ is the projected terrain elevation feature.

[0115] 2. Attention weight generation:

[0116] Dynamic weight allocation using a gating mechanism:

[0117]

[0118] Q2 = 1 - Q1;

[0119] Wherein, denotes vector concatenation, S ′ denotes the spectral reflectance feature, T ′ denotes the terrain elevation feature, W g ∈R 2d×1 is the gating weight matrix, that is, W g is a 2d row 1 column matrix, δ is the Sigmoid function, Q1 is the spectral feature weight, and Q2 is the terrain feature weight.

[0120] 3. Feature fusion:

[0121] F = Q1 ⊙ S ′ + Q2 ⊙ T ′ ;

[0122] Wherein, ⊙ represents element-wise multiplication.

[0123] In the present invention, through the gating weight matrix combined with the dynamic weight allocation function, automatic weight adjustment is achieved. For example, in a plain area (with a single terrain feature), the weight of the spectral feature is increased to more than 0.7, having good dynamic adaptability; the projection matrix can eliminate the dimensional difference between the spectral and terrain data, making the attention calculation more stable, and achieving feature decoupling; and the weight can be used as a decision basis (such as triggering a terrain anomaly review when Q1 < 0.3).

[0124] In order to generate an optimal ecological restoration plan, based on the above embodiments, in the embodiments of the present invention, the above multi-objective genetic algorithm is a parallel multi-objective genetic algorithm.

[0125] In order to generate an optimal ecological restoration plan, based on the above embodiments, in the embodiments of the present invention, with the ecological environment constraints as the constraint conditions and the optimal ecological restoration result as the objective function, the multi-objective genetic algorithm is used to solve the objective function, and the ecological restoration plan for the set area includes:

[0126] With the ecological environment constraints as the constraint conditions, the population individuals of the multi-objective genetic algorithm are initialized; the initialized population individuals include the initial ecological restoration plan;

[0127] Calculate the fitness values of the initialized population individuals according to the fitness function, perform multiple crossovers and mutations on the initialized population individuals to obtain new population individuals; iteratively update the population individuals until the maximum iteration termination condition is reached, and use the ecological restoration plan corresponding to the new population individuals at this time as the ecological restoration plan corresponding to the optimal ecological restoration result of the objective function.

[0128] The constraint conditions include vegetation coverage rate constraint, soil erosion constraint, biodiversity constraint, soil heavy metal content constraint value and surface runoff constraint.

[0129] In the specific implementation of the optimization and renovation based on the multi-objective genetic algorithm, the electronic device initializes the population individuals of the multi-objective genetic algorithm with the ecological environment restrictions as the constraint conditions; the initialized population individuals include the initial ecological restoration plan. This initial solution represents a set of possible ecological restoration plans, such as different vegetation planting densities, soil and water conservation project layouts, etc. Each solution is represented by a set of parameters, such as vegetation coverage rate, soil erosion index and biodiversity. Take the initialization result as the initial solution corresponding to the objective function, and optimize and iterate the initial solution based on the preset constraint conditions, and the constraint conditions include the minimum value of the vegetation coverage rate, the maximum value of the soil erosion index, the minimum value of the biodiversity, etc.

[0130] In the present invention, key ecological indicators such as vegetation coverage rate, soil erosion index and biodiversity can be extracted based on multi-source data and land types. The vegetation coverage rate can be calculated through the vegetation index (such as NDVI) in the hyperspectral image, the soil erosion index is estimated based on the terrain elevation and rainfall data, and the biodiversity is predicted through the species distribution model. These indicators are used as variables to construct the objective function. The form of the objective function is the weighted sum of the vegetation coverage rate, the soil erosion index and the biodiversity, and the weight coefficients are adjusted according to the importance of regional ecological protection.

[0131] In each iteration, calculate the fitness values of the initialized population individuals according to the fitness function, and obtain new population individuals through genetic operations such as selection, crossover and mutation. Among them, the selection operation is to screen high-quality solutions according to the objective function values, the crossover operation is to generate new solutions by combining the parameters of different solutions, and the mutation operation is to introduce diversity by randomly adjusting the parameters of the solutions.

[0132] Iteratively update the population individuals until the maximum iteration termination condition is reached. The algorithm can find a set of optimal solutions, and use the ecological restoration plan corresponding to the new population individuals corresponding to the optimal solutions as the ecological restoration plan corresponding to the optimal ecological restoration result of the objective function, that is, the optimal ecological restoration plan. The present invention can maximize the objective function value on the premise of meeting the constraint conditions, so as to achieve the optimal effect of ecological restoration.

[0133] For example, the constraints include vegetation coverage constraints, soil erosion constraints, biodiversity constraints, soil heavy metal content constraint values, and surface runoff constraints.

[0134] During the optimization and remediation process, the specific settings of the constraints include the minimum value of vegetation coverage, the maximum value of the soil erosion index, the minimum value of biodiversity, the maximum value of soil heavy metal content, and the maximum value of the surface runoff coefficient. For example, the minimum value of vegetation coverage is set to 30% to ensure that the vegetation coverage in the area meets the basic requirements of ecological protection; the maximum value of the soil erosion index is set to 0.5 to control the degree of soil erosion; the minimum value of biodiversity is set to 0.7 to protect the biodiversity in the area; the maximum value of soil heavy metal content is set to the national standard limit to prevent soil pollution; the maximum value of the surface runoff coefficient is set to 0.6 to reduce the flood risk. These constraints together constitute the boundary conditions for optimization and remediation, ensuring that the generated ecological restoration plan meets both ecological protection requirements and feasibility.

[0135] In another example, the variables of the objective function can be the intensity of the remediation project, the proportion of mixed sowing of vegetation, the density of terrace construction, etc. For example, the intensity of the remediation project is adjusted within the range of 0 - 100%.

[0136] For example, if the variables of the objective function are the intensity of the remediation project, the proportion of mixed sowing of vegetation, and the density of terrace construction, the electronic device can determine the optimal ecological restoration plan in the following ways:

[0137] 1. Construct an objective function with variables of the intensity of the remediation project, the proportion of mixed sowing of vegetation, and the density of terrace construction;

[0138] 2. Expand the constraints to soil heavy metal content and surface runoff coefficient; for example, expand the constraints to arsenic As ≤ 30 mg / kg and surface runoff coefficient < 0.45;

[0139] 3. Use the multi-objective genetic algorithm (Non-dominated Sorting Genetic Algorithm III, NSGA-III) to output the Pareto optimal solution set within 72 hours;

[0140] 4. Automatically optimize the algorithm weights according to the quarterly monitoring data, and prioritize improving the soil and water conservation rate.

[0141] Taking a specific example for illustration, the electronic device can determine the optimal ecological restoration plan through the following:

[0142] 1. Population initialization:

[0143] Coding rule: Real number coding is adopted. Each individual contains 3 groups of decision variables, namely, the treatment intensity, the vegetation mixed seeding ratio, and the terrace density. Among them, the treatment intensity is adjusted within the range of 0.0 - 1.0, and the adjustment step size is 0.05. The vegetation mixed seeding ratio is the ratio of herb, shrub, and arbor. The ratio of herb:shrub:arbor is adjusted within the range of 3:2:1 - 1:1:1. The terrace density is adjusted within the range of 5 - 20 strips / hectare.

[0144] Diversity control: Latin hypercube sampling is used to generate the initial population (scale ≥ 500), covering the entire feasible solution space.

[0145] 2. Crossover and mutation:

[0146] Directed crossover: For the top 10% of the individuals before non - dominated sorting, the simulated binary crossover (SBX) crossover operator is used (such as SBX crossover operator η = 20).

[0147] Adaptive mutation: The mutation rate is dynamically adjusted according to the degree of constraint violation (such as adjusting the mutation rate from 0.1 to 0.4), and polynomial mutation is used.

[0148] 3. Parallel computing architecture

[0149] Hierarchical parallelism:

[0150] The master node runs the NSGA - III framework, responsible for population sorting and elite retention.

[0151] The computing nodes use a hybrid programming of the Message Passing Interface (MPI) and Open Multi - Processing (OpenMP), that is, MPI + OpenMP hybrid programming, and calculate the fitness in blocks: Fitness = 0.6×soil and water conservation rate+0.3×vegetation coverage rate - 0.1×engineering cost. Among them, the number of computing nodes is not less than 8.

[0152] Dynamic load balancing: Re - allocate the computing tasks every 5 generations, and the latency difference < 50ms.

[0153] 4. Constraint handling mechanism:

[0154] Hard constraints, used to directly eliminate individuals:

[0155] Soil erosion modulus > 1500t / (km 2 ·a);

[0156] The proportion of heavy metal pollution area ≥ 5%;

[0157] Soft constraints, implemented based on the penalty function method:

[0158] Penalty item = max(0, surface runoff coefficient - 0.45)^2 × 100;

[0159] 5. Dynamic weight adjustment:

[0160] Adjust the target weight according to the restoration progress every 20 generations: If the vegetation survival rate < 60%, increase the weight of soil and water conservation rate to 0.7; If the project budget overspending > 15%, increase the cost penalty coefficient to 0.25.

[0161] In the embodiment of the present invention, there is no restriction on the termination condition of iteration, and it can be processed according to the iteration times reaching the maximum iteration times threshold.

[0162] In order to realize the supervision of ecological restoration in a set area and improve the accuracy of ecological restoration, after using the multi-objective genetic algorithm to solve the objective function with the above ecological environment constraints as the constraint conditions and the optimal ecological restoration result as the objective function to obtain the ecological restoration plan for the set area, it further includes:

[0163] Based on the ecological restoration plan, carry out ecological restoration on the set area;

[0164] Based on the multi-source monitoring data of the set area after ecological restoration, use the Bayesian network evaluation model to evaluate the restoration degree and obtain the ecological restoration degree;

[0165] Based on the ecological restoration degree, carry out ecological restoration supervision on the set area.

[0166] In the present invention, the electronic device can carry out ecological restoration on the set area based on the ecological restoration plan, and based on the multi-source monitoring data of the set area after ecological restoration, use the Bayesian network evaluation model to evaluate the restoration degree and obtain the ecological restoration degree, so as to realize the supervision of ecological restoration in the set area.

[0167] Among them, the ecological restoration degree can be the probability that the ecological index reaches the expected target.

[0168] Exemplarily, the ecological restoration degree determined by the Bayesian network evaluation model is that the probability of the vegetation coverage rate reaching the expected target is 85%, and the probability of the soil and water loss index decreasing to the expected target is 90%. If the restoration compliance rate does not reach the preset threshold (for example, 80%), then re-execute the step of generating the optimal ecological restoration plan using the multi-objective genetic algorithm based on the target map. In the present invention, the electronic device can continuously optimize the ecological restoration plan to ensure that the restoration effect reaches the expected target.

[0169] Exemplarily, the Bayesian network evaluation model is based on 12-dimensional ecological indicators (including avian diversity index and groundwater pH value), outputs the degree of ecological restoration (0 - 1), and dynamically feeds back to the optimization module to form a closed loop, increasing the restoration compliance rate from 65% to 89%.

[0170] In addition, a microbial activity index evaluation system can be constructed to conduct effect evaluation after the optimal ecological restoration plan has been running for a period of time. Specifically, by deploying a micro-biosensor network, the soil ATP concentration is monitored in real time. For example, the real-time monitoring of the soil ATP concentration is not less than 1 nmol / g, and the litter decomposition rate is retrieved by combining the multi-spectral data of drones, so as to conduct effect evaluation.

[0171] The present invention solves the problem of lag in traditional regulatory data through multi-source data fusion and an adaptive prediction model; adopts a dynamic evaluation index and an optimized closed loop for the restoration plan, increasing the ecological restoration efficiency by more than 30%; combines the early warning mechanism with three-dimensional visualization to achieve a full-process response speed of "monitoring - early warning - disposal" < 2 hours.

[0172] The following uses a specific embodiment to illustrate the embodiments of the present invention. In this embodiment, the following steps are included:

[0173] (1) Obtain the hyperspectral image and lidar point cloud data of the area to be supervised.

[0174] (2) Extract features from the hyperspectral image to obtain spectral reflectance features; extract features from the lidar point cloud data to obtain terrain elevation features; use a cross-modal attention gating mechanism to assign weights to the spectral reflectance features and terrain elevation features to obtain spectral feature weights and terrain feature weights; based on the spectral reflectance features, terrain elevation features, spectral feature weights and terrain feature weights, use the output layer of the multi-modal feature fusion network to conduct land use classification to obtain the land types of the set area.

[0175] (3) Based on the multi-source monitoring data and land types, use the dynamic prediction sub-module to predict land use changes to obtain the land use change prediction results.

[0176] (4) If the land use change prediction result exceeds the land development early warning range, then use the ecological environment restriction as a constraint condition and the optimal ecological restoration result as the objective function, and use the multi-objective genetic algorithm to solve the objective function to obtain the ecological restoration plan for the set area.

[0177] (5) Based on the ecological restoration plan, conduct ecological restoration on the set area; based on the multi-source monitoring data of the set area after ecological restoration, use the Bayesian network evaluation model to evaluate the restoration degree to obtain the ecological restoration degree; based on the ecological restoration degree, conduct ecological restoration supervision on the set area.

[0178] In the present invention, hyperspectral images provide rich spectral information and can accurately identify the spectral characteristics of different landforms. LiDAR point cloud data provides high-precision terrain information and can reflect the subtle undulations of the surface. The present invention uses a multimodal feature fusion network to conduct a comprehensive analysis of hyperspectral images and LiDAR point cloud data to obtain land types, so that land use can be predicted for a set area based on the land type, and land use change prediction results can be obtained. If the land use change prediction result exceeds the land development warning range, the ecological environment restriction is used as a constraint condition, and the optimal ecological restoration result is used as the objective function. The objective function is solved using a multi-objective genetic algorithm to obtain an ecological restoration plan for the set area, so that the optimal restoration plan can be found under complex ecological constraint conditions, significantly improving the restoration effect and resource utilization efficiency. This method significantly improves the restoration efficiency and reduces the restoration cost, provides a sustainable solution for regional ecological protection, and enhances the robustness and adaptability of the system.

[0179] This invention achieves comprehensive oversight of land consolidation across the entire region through multi-source remote sensing data acquisition, real-time monitoring and analysis, land use change prediction, and ecological restoration effectiveness assessment modules. It includes an unmanned aerial vehicle (UAV) image acquisition unit, an AI-powered intelligent analysis module, a dynamic data visualization platform, and an early warning and decision support system. Using intelligent algorithms, it achieves land use classification, optimizes restoration targets, and provides real-time early warnings. This addresses existing issues such as low land consolidation supervision efficiency and delayed data updates, providing efficient decision-making support for governments and relevant departments.

[0180] Example 2:

[0181] Based on the same inventive concept, the present invention also provides a smart national land space ecological supervision and restoration system, the structural diagram of which is as follows: Figure 2 Shown, including:

[0182] Acquisition module 201, used to acquire multi-source monitoring data of a set area;

[0183] The land use prediction module 202 is used to predict land use in a set area based on multi-source monitoring data using a dynamic prediction model to obtain land use change prediction results;

[0184] The ecological restoration module 203 is used to use the ecological environment restrictions as constraints and the optimal ecological restoration results as the objective function if the land use change prediction results exceed the land development warning range, and use a multi-objective genetic algorithm to solve the objective function to obtain an ecological restoration plan for the set area.

[0185] In a specific implementation manner, the land use prediction module 202 is specifically configured to perform land use classification based on multi-source monitoring data by using a multi-modal feature fusion network to obtain the land types of a set area; and perform land use change prediction based on the multi-source monitoring data and the land types by using a dynamic prediction sub-module to obtain a land use change prediction result.

[0186] In a specific implementation manner, the land use prediction module 202 is specifically configured to use the three-dimensional convolutional layer of the multi-modal feature fusion network to extract features from the hyperspectral image and lidar point cloud data in the multi-source monitoring data to obtain spatial-spectral joint features; the spatial-spectral joint features include spectral reflectance features and terrain elevation features; use the cross-modal attention gating mechanism of the multi-modal feature fusion network to assign weights to the spectral reflectance features and terrain elevation features to obtain spectral feature weights and terrain feature weights; and perform land use classification based on the spectral reflectance features, terrain elevation features, spectral feature weights, and terrain feature weights by using the output layer of the multi-modal feature fusion network to obtain the land types of a set area.

[0187] In a specific implementation manner, the land use prediction module 202 is specifically configured to use the cross-modal attention gating mechanism of the multi-modal feature fusion network and combine it with a dynamic weight assignment function to assign weights to the spectral reflectance features and terrain elevation features to obtain spectral feature weights and terrain feature weights;

[0188] The dynamic weight assignment function satisfies the following formula:

[0189]

[0190] Q2 = 1 - Q1;

[0191] Where, denotes vector concatenation, S ′ denotes the spectral reflectance feature, T ′ denotes the terrain elevation feature, W g is the gating weight matrix, δ is the Sigmoid function, Q1 is the spectral feature weight, and Q2 is the terrain feature weight.

[0192] In a specific implementation manner, the multi-objective genetic algorithm is a parallel multi-objective genetic algorithm.

[0193] In a specific implementation manner, the ecological restoration module 203 is specifically configured to initialize the population individuals of the multi-objective genetic algorithm with the ecological environment constraints as the constraint conditions; the initialized population individuals include the initial ecological restoration plan; calculate the fitness values of the initialized population individuals according to the fitness function, perform multiple crossovers and mutations on the initialized population individuals to obtain new population individuals; iteratively update the population individuals until the maximum iteration termination condition is reached, and use the ecological restoration plan corresponding to the new population individuals at this time as the ecological restoration plan corresponding to the optimal ecological restoration result of the objective function; the constraint conditions include vegetation coverage constraint, soil erosion constraint, biodiversity constraint, soil heavy metal content constraint value, and surface runoff constraint.

[0194] In a specific implementation manner, the device further includes:

[0195] The ecological restoration supervision module 204 is configured to perform ecological restoration on the set area based on the ecological restoration plan; evaluate the restoration degree using the Bayesian network evaluation model based on the multi-source monitoring data of the set area after ecological restoration to obtain the ecological restoration degree; perform ecological restoration supervision on the set area based on the ecological restoration degree.

[0196] Embodiment 3:

[0197] As Figure 3 shown, the present invention further provides an electronic device, which may be a computer device, a single-chip microcomputer device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, the processor, and the transceiver component are connected through a bus; the memory can be used to store an execution program, and an exemplary execution program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, and the data can be called and / or modified when the instructions are executed.

[0198] The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a land ecological restoration method in the above embodiment.

[0199] Example 4:

[0200] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device-readable storage medium (Memory). The electronic device-readable storage medium is a memory device in the electronic device, used to store programs and data. It can be understood that the storage medium here can include both the built-in storage medium in the electronic device and, of course, the extended storage medium supported by the electronic device. The storage medium provides a storage space, and this storage space stores the operating system of the terminal. And, in this storage space, there is also stored one or more instructions suitable for being loaded and executed by the processor. These instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. By the processor loading and executing one or more instructions stored in the storage medium, the steps of a method for land ecological restoration in the above embodiments can be realized.

[0201] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0202] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0203] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and this instruction device realizes the functions in the flowFigure 1 one process or multiple processes and / or boxes Figure 1 the functions specified in one box or multiple boxes

[0204] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or boxes Figure 1 one box or multiple boxes

[0205] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the scope of its protection. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that after reading the present invention, various changes, modifications or equivalent replacements can still be made to the specific implementation manners of the application. However, these changes, modifications or equivalent replacements are all within the scope of protection of the claims pending for approval of the application.

Claims

1. A method for land ecological restoration, characterized in that, Including: Obtain multi-source monitoring data of a set area; Based on the multi-source monitoring data, use a dynamic prediction model to predict land use in the set area, and obtain a land use change prediction result; If the land use change prediction result exceeds the land development warning range, use the ecological environment restriction as a constraint condition and the optimal ecological restoration result as an objective function, and use a multi-objective genetic algorithm to solve the objective function to obtain an ecological restoration plan for the set area.

2. The method according to claim 1, characterized in that The step of using a dynamic prediction model to predict land use in the set area based on the multi-source monitoring data to obtain a land use change prediction result includes: Based on the multi-source monitoring data, use the multi-modal feature fusion network in the dynamic prediction model to classify land use and obtain the land type of the set area; Based on the multi-source monitoring data and the land type, use the dynamic prediction sub-module in the dynamic prediction model to predict land use change and obtain a land use change prediction result.

3. The method according to claim 2, characterized in that The step of using the multi-modal feature fusion network in the dynamic prediction model to classify land use based on the multi-source monitoring data to obtain the land type of the set area includes: Use the 3D convolutional layer of the multi-modal feature fusion network in the dynamic prediction model to extract features from the hyperspectral image and lidar point cloud data in the multi-source monitoring data, and obtain a spatial-spectral joint feature; the spatial-spectral joint feature includes a spectral reflectance feature and a terrain elevation feature; Use the cross-modal attention gating mechanism of the multi-modal feature fusion network to assign weights to the spectral reflectance feature and the terrain elevation feature, and obtain a spectral feature weight and a terrain feature weight; Based on the spectral reflectance feature, the terrain elevation feature, the spectral feature weight, and the terrain feature weight, use the output layer of the multi-modal feature fusion network to classify land use and obtain the land type of the set area.

4. The method according to claim 3, characterized in that The step of using the cross-modal attention gating mechanism of the multi-modal feature fusion network to assign weights to the spectral reflectance feature and the terrain elevation feature to obtain a spectral feature weight and a terrain feature weight includes: Use the cross-modal attention gating mechanism of the multi-modal feature fusion network and combine it with a dynamic weight assignment function to assign weights to the spectral reflectance feature and the terrain elevation feature, and obtain a spectral feature weight and a terrain feature weight; The dynamic weight assignment function satisfies the following formula: Q2 = 1 - Q1; in, Represents vector concatenation, S ′ represents the spectral reflectance characteristics, T ′ Represents the terrain elevation feature, W g is the gating weight matrix, δ is the Sigmoid function, Q1 is the spectral feature weight, and Q2 is the terrain feature weight.

5. The method according to claim 1, characterized in that The multi-objective genetic algorithm is a parallel multi-objective genetic algorithm.

6. The method according to claim 1 or 5, characterized in that, The step of using a multi-objective genetic algorithm to solve the objective function with the ecological environment restriction as a constraint condition and the optimal ecological restoration result as an objective function to obtain an ecological restoration plan for the set area includes: Use the ecological environment restriction as a constraint condition to initialize the population individuals of the multi-objective genetic algorithm; the initialized population individuals include an initial ecological restoration plan; Calculate the fitness values of the initialized population individuals according to the fitness function, perform multiple crossovers and mutations on the initialized population individuals to obtain new population individuals; iteratively update the population individuals until the maximum iteration termination condition is reached, and use the ecological restoration plan corresponding to the new population individuals at this time as the ecological restoration plan corresponding to the optimal ecological restoration result of the objective function. The constraint conditions include vegetation coverage constraint, soil erosion constraint, biodiversity constraint, soil heavy metal content constraint value, and surface runoff constraint.

7. The method according to claim 1, characterized in that After using the multi-objective genetic algorithm to solve the objective function with the ecological environment limit as the constraint condition and the optimal ecological restoration result as the objective function to obtain the ecological restoration plan for the set area, it further includes: Based on the ecological restoration plan, carry out ecological restoration on the set area. Based on the multi-source monitoring data of the set area after ecological restoration, use the Bayesian network evaluation model to evaluate the restoration degree to obtain the ecological restoration degree. Based on the ecological restoration degree, conduct ecological restoration supervision on the set area.

8. A land ecological restoration system, characterized in that: It includes: An acquisition module for acquiring multi-source monitoring data of a set area. A land use prediction module for predicting the land use of the set area using a dynamic prediction model based on the multi-source monitoring data to obtain a land use change prediction result. An ecological restoration module for, if the land use change prediction result exceeds the land development warning range, using the multi-objective genetic algorithm to solve the objective function with the ecological environment limit as the constraint condition and the optimal ecological restoration result as the objective function to obtain the ecological restoration plan for the set area.

9. An electronic device, characterized in that, It includes: At least one processor and a memory. The memory and the processor are connected by a bus. The memory is used to store one or more programs. When the one or more programs are executed by the at least one processor, the land ecological restoration method described in any one of claims 1-7 is implemented.

10. A readable storage medium, characterized in that, There is an execution program stored thereon, and when the execution program is executed, the land ecological restoration method described in any one of claims 1-7 is implemented.

Images