Mine ecological safety identification method based on convolutional neural network model

Abstract

The invention discloses a mine ecological safety identification method based on a convolutional neural network model, and relates to the technical field of mine ecological safety identification. A convolutional neural network CNN is introduced to automatically optimize a resistance value, a time dimension and a dynamic factor are combined, and a resistance surface is monitored and adjusted in real time, so that the safety of mine ecological safety identification is improved. The method can dynamically respond to the influence of topographic change, vegetation density fluctuation and meteorological change, can capture the real-time influence of sudden rainfall, mining area extension or vegetation restoration and the like, the generated resistance surface has high timeliness and accurately reflects the dynamic characteristics of an ecological system, and meanwhile, the circuit theory and deep learning are combined, so that the method can be widely applied to the field of ecological engineering. According to the method, multiple potential migration paths are generated, dependence on a single minimum resistance path is avoided, more diversified migration choices are provided, path weights are dynamically optimized in combination with a time sequence model, channels with high ecological values and high migration requirements are preferentially protected, and it is guaranteed that a migration network has enough elasticity and adaptability.

Description

Technical Field

[0001] The present invention relates to the technical field of mine ecological safety identification, and in particular to a mine ecological safety identification method based on a convolutional neural network model. Background Art

[0002] In the identification of mine ecological safety, traditional methods mostly use static ecological resistance surfaces and single-path minimum cumulative resistance models to identify ecological corridors and key ecological nodes. They rely on remote sensing images or geographic information system (GIS) data to construct resistance surfaces, and map different land use types, terrain, vegetation coverage and other factors into fixed resistance values, thereby describing migration pathways in the ecosystem.

[0003] However, the mining environment is greatly affected by human mining, and ecological factors such as surface morphology and vegetation coverage change significantly over time, making it difficult for traditional static resistance surfaces to accurately reflect the dynamic characteristics of mining ecology, and their timeliness and applicability in ecological identification are low.

[0004] To address this problem, some traditional solutions increase the monitoring frequency and zoning, identify and set up ecological corridors in advance to cope with changes in the ecological environment. Specific methods include setting up ecological corridors in major ecologically sensitive areas, areas with good vegetation coverage, or key species activity areas, and adjusting the resistance surface based on the current situation; in addition, some solutions improve the model and increase the number of pathways. However, due to the complexity and dynamic nature of the mining area ecosystem, such static solutions are difficult to respond in real time, and human intervention is usually required when adjusting the corridor, which cannot achieve real-time updates and true adaptability. Therefore, there is an urgent need to use a mine ecological safety identification method based on a convolutional neural network model to solve such problems. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a mine ecological safety identification method based on a convolutional neural network model to solve the problem that static resistance surfaces and fixed single-path models cannot adapt to rapid environmental changes, resulting in insufficient connectivity identification or large deviations, and cannot accurately reflect the diversity of species migration paths; traditional methods are difficult to achieve dynamic updating and multi-path identification of ecological corridors, and lack real-time and high-resolution response mechanisms.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] The present invention provides a mine ecological safety identification method based on a convolutional neural network model, which comprises: step S1, mine data collection,

[0009] Collect mine data, including topographic data, vegetation density, meteorological data, and mining record data;

[0010] Step S2, dynamic resistance surface modeling,

[0011] Based on the mine data collected in step S1, a dynamic ecological resistance surface is constructed to reflect the spatial distribution of the degree of obstruction of species migration in each region of the ecosystem, and the resistance value of each region depends on the ecological factors;

[0012] Step S3, multi-path ecological corridor identification,

[0013] Using a combination of deep learning and circuit theory, multiple ecological corridors, i.e. possible migration paths, are identified and generated on the ecological resistance surface. This is not limited to the path of least resistance, but also integrates ecosystem factors, including species migration needs and resource distribution.

[0014] In step S3, based on the deep learning model combined with time series analysis, the priority of each path is adjusted to give priority to protecting corridors with high ecological value and high-demand species passages;

[0015] Step S4: Dynamic identification of ecological pinch points and obstacle points.

[0016] Based on the ecological resistance surface constructed in step S2, identify pinch points and obstacle points in the ecosystem, i.e., bottlenecks and blocking areas in the migration path;

[0017] Step S5: Ecological management based on dynamic results.

[0018] Based on the ecological corridors, pinch points and obstacle points identified in step S3 and step S4, targeted ecological protection and restoration measures are taken; the identification results are transmitted to the ecological management system in real time, and the mine management can give priority to repairing or protecting key corridors and pinch points to reduce the impact of human activities on ecological connectivity.

[0019] Furthermore, in step S2, a convolutional neural network (CNN) is used to dynamically adjust the resistance value according to the terrain data, vegetation density, and water distribution in the meteorological data to simulate the obstruction of the mining ecosystem to species migration.

[0020] Furthermore, in step S2, based on the data collected in step S1, a convolutional neural network (CNN) is used to dynamically adjust the resistance value. The specific steps include:

[0021] The terrain data (slope T(x, y)), vegetation density V(x, y) and meteorological data (water distribution W(x, y)) are spatially rasterized:

[0022] R raw (x, y) = f norm (T(x,y))+gnorm (V(x,y))+h norm (W(x, y)), where R raw (x, y) represents the initial resistance value of the area (x, y), f norm (·), g norm (·), h norm (·) represents the function for normalizing each ecological factor, keeping the value range between [0, 1]. T(x, y) represents the terrain slope value of the region (x, y). The greater the slope, the higher the resistance. V(x, y) represents the vegetation coverage density of the region (x, y). The higher the coverage, the lower the resistance. W(x, y) represents the water distribution characteristics of the region (x, y). The closer to the water body, the lower the resistance.

[0023] Based on the initial resistance value R raw (x, y), and dynamically adjust it using a convolutional neural network (CNN), capturing local spatial characteristics and reflecting the interactive relationship of ecological factors, and dynamically adjusting resistance:

[0024] Among them, R dyn (x, y) represents the dynamically adjusted resistance value of the region (x, y), σ(·) represents the ReLU activation function, which is used to maintain the nonlinear characteristics of the output, and K i，j Represents the weight matrix of the convolution kernel, representing the weights of different positions, k represents the size of the convolution kernel, b represents the bias term, which is used to dynamically adjust the baseline of the overall resistance value, R raw (x+i, y+j) represents the local neighborhood pixel value of the initial resistance value.

[0025] Furthermore, in step S2, an environmental change response mechanism is introduced:

[0026] Introducing the time dimension t and the dynamic factor D(x, y, t), capturing the impact of environmental changes based on the time series, and calculating the dynamic resistance value of the time dimension:

[0027] R final (x, y, t) = R dyn (x, y)·(1+α·D(x, y, t)), where, R final (x, y, t) represents the final dynamic resistance value of the area (x, y) at time t, α represents the weight coefficient of the dynamic factor, which is used to adjust the influence of environmental changes on the resistance value, and D(x, y, t) represents the dynamic environmental factor (such as sudden rainfall or vegetation restoration) of the area (x, y) at time t, which is obtained from measured data and ranges from [-1, 1] to generate the resistance surface map: in, represents the dynamic ecological resistance surface at time t, Ω represents the research scope of the ecosystem, i.e., the total area, and Rfinal (x, y, t) represents the final resistance value for each pixel.

[0028] Furthermore, in step S3, multi-path identification of ecological corridors is performed.

[0029] In the dynamic ecological resistance On the other hand, circuit theory is used to calculate the flow of current when species migrate. Among them, I(x, y) represents the current intensity of pixel (x, y), which represents the flow intensity in the ecological migration network. represents the dynamic ecological resistance value generated in step S2, S and T represent the migration starting point set and target point set in the ecosystem, V s , V t It represents the voltage values ​​of the starting point s and the target point t, which are set to 1 and 0 to simulate the migration path selection of species on the ecological resistance surface. The area with high current intensity is a potential corridor;

[0030] Using the current flow model I(x, y), potential migration paths are identified: P k (x, y) = H (I (x, y) - θ k ), P k (x, y) represents the indication value of path k at pixel (x, y), 1 means the path passes through the pixel, 0 means it does not pass through, H(·) represents the unit step function, when the input is greater than 0, the output is 1, otherwise it outputs 0, θ k Represents the current threshold of path k, which is used to control the path flow, and the control threshold θ k Adjust the number of paths to generate multiple evenly distributed migration paths; introduce a deep learning model to dynamically optimize path weights based on ecological factors:

[0031] in, represents the optimization objective function, which is the total resistance and negative impact of all paths, N represents the total number of candidate migration paths, λ1, λ2, λ3 represent weight parameters, which measure the influence of resistance, terrain gradient and dynamic factors respectively. Represents the terrain gradient of pixel (x, y), corresponding to the slope cost of migration, and D(x, y, t) represents the dynamic environmental factors of pixel (x, y) at time t, such as the impact of rainfall. Through the deep learning framework, the path distribution is optimized to make it more concentrated in the low-resistance area, while reducing the negative impact of the dynamic environment on the path.

[0032] Furthermore, in step S3, the path priority is adjusted based on the time series:

[0033] Combine the time series model to predict the importance of each path and dynamically adjust the priority:

[0034] Among them, W k (t) represents the priority of path k at time t, β1, β2 represent the priority adjustment weights, and H k (t) represents the dynamic demand index of path j, the species migration demand based on time series prediction, and dxdy represents the integration unit in two-dimensional space;

[0035] Use time series analysis to predict future migration demand:

[0036] H k (t+1) = γ·H k (t)+δ·Env(t), where H k (t+1) represents the migration demand of path k at time t+1, γ represents the time attenuation factor of the migration demand, δ represents the response coefficient of environmental change, Env(t) represents the real-time change of environmental factors at time t,

[0037] Combined with the path priority, select high ecological value paths for protection: P prior = argmax k W k (t), where P prior represents the set of paths to be protected first, W k (t) represents the current priority of path k, dynamically selecting migration paths with high ecological value and high demand for priority protection and restoration.

[0038] Furthermore, in step S4, based on the circuit theory model, the current density in different regions is calculated to identify high-importance nodes in the ecological network, i.e., pinch points, and obstacle points that hinder species migration.

[0039] Further, in step S4, the positions and priorities of the pinch points and the obstacle points are adjusted according to the changes in mining, including the expansion of the mining area and the destruction of vegetation;

[0040] In step S4, the vegetation density changes and meteorological changes collected in step S1 are combined, and the agent-based simulation method is used to simulate the frequency and pressure of different species passing through pinch points and obstacle points in the channel, and the positions of pinch points and obstacle points are dynamically adjusted.

[0041] Furthermore, in step S4, the current density is calculated based on the resistance surface:

[0042] Using circuit theory, the current distribution is simulated on the ecological resistance surface and the current density in each area is calculated:

[0043] Where J(x, y) represents the current density of pixel (x, y), that is, the pressure level of the point as a migration path in the ecological network, S, T represent the starting point set S and the end point set T, respectively defining the source and target area of ​​migration, V s , V t represents the voltage value at the starting point s and the end point t, G(x, y) represents the conduction rate of the pixel (x, y), which is defined as Inversely proportional to the resistance value, represents the resistance value of pixel (x, y) on the ecological resistance surface. Here, the current density J(x, y) reflects the importance of the connectivity of pixel (x, y) in the migration network in the ecosystem. A high current density indicates a bottleneck position in the path that is critical to migration.

[0044] Based on the current density map, identify the current high value concentration area as the pinch point: C(x, y) = I(J(x, y)>θ J ), where C(x, y) indicates whether the pixel (x, y) is a pinch point, 1 indicates a pinch point, 0 indicates a non-pinch point, I(·) indicates an indicator function, outputs 1 when the condition is met, otherwise outputs 0, θ J represents the current density threshold, which is used to define the importance of the pinch point. The area with current density higher than the threshold is marked as the pinch point, which represents the key bottleneck point in the migration path;

[0045] Conductivity and resistance change rate are used to detect obstacle points, i.e., blocked areas:

[0046] Among them, B(x, y) indicates whether the pixel (x, y) is an obstacle point, 1 indicates an obstacle point, and 0 indicates a non-obstacle point. Indicates the resistance value change rate of pixel (x, y), indicating the dynamic change trend of the degree of obstruction, θ B It represents the resistance change rate threshold, which is used to define the criteria for determining obstacle points. Areas where the resistance value change rate is higher than the threshold are marked as obstacle points, representing dynamically changing areas that may hinder species migration.

[0047] Furthermore, in step S4, the pinch points and the obstacle points are adjusted dynamically according to the mining conditions:

[0048] According to the mining expansion area, the current density threshold is redefined: θ' J (t) = θ J ·(1+α·A mine (t)), where θ' J (t) represents the pinch current density threshold after dynamic adjustment at time t, α represents the influence coefficient of mine expansion on current density, A mine(t) represents the proportion of the mine expansion area at time t. As the mine expansion area increases, the current density threshold is increased to avoid misjudging non-critical nodes as pinch points;

[0049] According to the dynamic changes of vegetation destruction, the obstacle point judgment condition is adjusted, θ' B (t) = θ B ·(1+β·L veg (t)), where θ' B (t) represents the resistance change rate threshold after dynamic adjustment at time t, β represents the influence coefficient of vegetation destruction on the resistance change rate, L veg (t) represents the vegetation coverage loss rate at time t. The higher the vegetation destruction rate, the lower the judgment threshold and the more sensitive the identification of possible obstacles.

[0050] In step S4, the pinch points and obstacle points are dynamically adjusted based on the agent simulation:

[0051] Agent-Based Modeling is used to simulate the frequency of species passing through pinch points in migration corridors:

[0052] Among them, F C (t) represents the frequency of agents passing through the pinch point C at time t, N A represents the total number of agents participating in the migration in the simulation, A i (t) represents the position of agent i at time t, I(A i (t)∈C) means that if agent i passes through pinch point C, output 1, otherwise output 0. The frequency of the pinch point reflects the importance of the point in the migration network. The pinch points with high frequency need to be protected first;

[0053] Simulate the stress level of the agent passing through the obstacle point: Among them, P B (t) represents the agent pressure level at obstacle point B at time t, It represents the resistance value at the location of agent i. The pressure level of the obstacle point reflects its negative impact on the migration path. Obstacle points with high pressure need to be repaired or alleviated first.

[0054] The beneficial effects of the present invention are:

[0055] The present invention introduces a convolutional neural network (CNN) to automatically optimize the resistance value, combines the time dimension and dynamic factors, and adjusts the resistance surface through real-time monitoring, so that it can dynamically respond to the influence of terrain changes, vegetation density fluctuations and meteorological changes, and can capture the real-time impact of changes such as sudden rainfall, mining area expansion or vegetation restoration. The generated resistance surface has high timeliness and accurately reflects the dynamic characteristics of the ecosystem.

[0056] The present invention combines circuit theory with deep learning to generate multiple potential migration paths, avoid relying on a single path of least resistance, provide more diverse migration options, and dynamically optimize path weights in combination with time series models, giving priority to protecting channels with high ecological value and high migration demand, ensuring that the migration network has sufficient flexibility and adaptability.

[0057] The present invention, through the combination of deep learning and circuit theory, realizes the automatic identification and dynamic adjustment of ecological resistance surfaces, migration paths, pinch points and obstacle points. It can adapt to the complex changes of the ecological environment of mines without human intervention, and uses agent simulation methods to quantify the dynamic pressure levels of pinch points and obstacle points, making protection and restoration decisions more accurate and more adaptable.

[0058] The present invention combines real-time monitoring data, and the model can automatically and dynamically adjust the resistance value and corridor layout, while adaptively optimizing the priorities of pinch points and obstacle points, greatly improving the adjustment efficiency. By transmitting the identification results to the ecological management system in real time, mine workers can quickly obtain dynamic adjustment suggestions and improve management efficiency.

[0059] The present invention adopts an agent simulation method that can simulate the behavior patterns of different species in migration channels. By quantitatively analyzing the pinch point frequency and obstacle point pressure, it ensures that the migration channels can meet the ecological needs of multiple species. Based on the simulation results, the channels of high-demand species are given priority protection to ensure the diversity and sustainability of the ecological network. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0061] Figure 1 It is a schematic diagram of the flow of the mine ecological safety identification method based on the convolutional neural network model of the present invention. DETAILED DESCRIPTION

[0062] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0063] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0064] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0065] Example 1, reference Figure 1 This embodiment provides a mine ecological safety identification method based on a convolutional neural network model, comprising the following steps:

[0066] Step S1, mine data collection,

[0067] Collect mine data, including topographic data, vegetation density, meteorological data, and mining record data;

[0068] Step S2, dynamic resistance surface modeling,

[0069] Based on the mine data collected in step S1, a dynamic ecological resistance surface is constructed to reflect the spatial distribution of the degree of obstruction of species migration in each region of the ecosystem, and the resistance value of each region depends on the ecological factors;

[0070] In step S2, a convolutional neural network (CNN) is used to dynamically adjust the resistance value according to the terrain data, vegetation density, and water distribution in the meteorological data to simulate the obstruction of the mining ecosystem to species migration;

[0071] In step S2, based on the data collected in step S1, a convolutional neural network (CNN) is used to dynamically adjust the resistance value. The specific steps include:

[0072] The terrain data (slope T(x, y)), vegetation density V(x, y) and meteorological data (water distribution W(x, y)) are spatially rasterized:

[0073] R raw (x, y) = f norm (T(x,y))+g norm (V(x,y))+h norm (W(x, y)), where R raw (x, y) represents the initial resistance value of the area (x, y), f norm (·), g norm (·), h norm(·) represents the function for normalizing each ecological factor, keeping the value range between [0, 1]. T(x, y) represents the terrain slope value of the region (x, y). The greater the slope, the higher the resistance. V(x, y) represents the vegetation coverage density of the region (x, y). The higher the coverage, the lower the resistance. W(x, y) represents the water distribution characteristics of the region (x, y). The closer to the water body, the lower the resistance.

[0074] Based on the initial resistance value R raw (x, y), and dynamically adjust it using a convolutional neural network (CNN), capturing local spatial characteristics and reflecting the interactive relationship of ecological factors, and dynamically adjusting resistance:

[0075] Among them, R dyn (x, y) represents the dynamically adjusted resistance value of the region (x, y), σ(·) represents the ReLU activation function, which is used to maintain the nonlinear characteristics of the output, and K i，j Represents the weight matrix of the convolution kernel, representing the weights of different positions, k represents the size of the convolution kernel, b represents the bias term, which is used to dynamically adjust the baseline of the overall resistance value, R raw (x+i, y+j) represents the local neighborhood pixel value of the initial resistance value;

[0076] In step S2, an environmental change response mechanism is introduced:

[0077] Introducing the time dimension t and the dynamic factor D(x, y, t), capturing the impact of environmental changes based on the time series, and calculating the dynamic resistance value of the time dimension:

[0078] R final (x, y, t) = R dyn (x, y)·(1+α·D(x, y, t)), where, R final (x, y, t) represents the final dynamic resistance value of the area (x, y) at time t, α represents the weight coefficient of the dynamic factor, which is used to adjust the influence of environmental changes on the resistance value, and D(x, y, t) represents the dynamic environmental factor (such as sudden rainfall or vegetation restoration) of the area (x, y) at time t, which is obtained from measured data and ranges from [-1, 1] to generate the resistance surface map: in, represents the dynamic ecological resistance surface at time t, Ω represents the research scope of the ecosystem, i.e., the total area, and R final (x, y, t) represents the final resistance value of each pixel;

[0079] Specifically, space and time are combined to introduce the time dimension into the resistance surface modeling. The dynamic factor D(x, y, t) is used to reflect the sudden changes in the environment, such as the expansion of mining or vegetation restoration, to enhance the timeliness of the resistance surface. i，j Automatic learning captures the interactive relationship between terrain, vegetation and water distribution, so that the resistance surface reflects the characteristics of the ecosystem more accurately. The dynamic response mechanism is introduced, combined with real-time monitoring data D(x, y, t), and the resistance value is automatically and dynamically adjusted without human intervention. The generated spatial distribution map Dynamically reflects migration resistance in the ecological environment;

[0080] Step S3, multi-path ecological corridor identification,

[0081] Using a combination of deep learning and circuit theory, multiple ecological corridors, i.e. possible migration paths, are identified and generated on the ecological resistance surface. This is not limited to the path of least resistance, but also integrates ecosystem factors, including species migration needs and resource distribution.

[0082] In step S3, based on the deep learning model combined with time series analysis, the priority of each path is adjusted to give priority to protecting corridors with high ecological value and high-demand species passages;

[0083] In step S3, multi-path identification of ecological corridors is performed.

[0084] In the dynamic ecological resistance On the other hand, circuit theory is used to calculate the flow of current when species migrate. Among them, I(x, y) represents the current intensity of pixel (x, y), which represents the flow intensity in the ecological migration network. represents the dynamic ecological resistance value generated in step S2, S and T represent the migration starting point set and target point set in the ecosystem, V s , V t It represents the voltage values ​​of the starting point s and the target point t, which are set to 1 and 0 to simulate the migration path selection of species on the ecological resistance surface. The area with high current intensity is a potential corridor;

[0085] Using the current flow model I(x, y), potential migration paths are identified: P k (x, y) = H (I (x, y) - θ k ), P k (x, y) represents the indication value of path k at pixel (x, y), 1 means the path passes through the pixel, 0 means it does not pass through, H(·) represents the unit step function, when the input is greater than 0, the output is 1, otherwise it outputs 0, θ k Represents the current threshold of path k, which is used to control the path flow, and the control threshold θ kAdjust the number of paths to generate multiple evenly distributed migration paths; introduce a deep learning model to dynamically optimize path weights based on ecological factors:

[0086] in, represents the optimization objective function, which is the total resistance and negative impact of all paths, N represents the total number of candidate migration paths, λ1, λ2, λ3 represent weight parameters, which measure the influence of resistance, terrain gradient and dynamic factors respectively. represents the terrain gradient of pixel (x, y), corresponding to the slope cost of migration, and D(x, y, t) represents the dynamic environmental factors of pixel (x, y) at time t, such as the impact of rainfall. Through the deep learning framework, the path distribution is optimized to make it more concentrated in the low-resistance area, while reducing the negative impact of the dynamic environment on the path;

[0087] In step S3, the path priority is adjusted based on the time series:

[0088] Combine the time series model to predict the importance of each path and dynamically adjust the priority:

[0089] Among them, W k (t) represents the priority of path k at time t, β1, β2 represent the priority adjustment weights, and H k (t) represents the dynamic demand index of path k, the species migration demand based on time series prediction, and dxdy represents the integration unit in two-dimensional space;

[0090] Use time series analysis to predict future migration demand:

[0091] H k (t+1) = γ·H k (t)+δ·Env(t), where H k (t+1) represents the migration demand of path k at time t+1, γ represents the time attenuation factor of the migration demand, δ represents the response coefficient of environmental change, Env(t) represents the real-time change of environmental factors at time t,

[0092] Combined with the path priority, select high ecological value paths for protection: P prior = argmax k W k (t), where P prior represents the set of paths to be protected first, W k (t) represents the current priority of path k, dynamically selecting migration paths with high ecological value and high demand for priority protection and restoration;

[0093] Specifically, the migration paths are identified through circuit theory, path characteristics are optimized through deep learning, and path priorities are adjusted through time series, ultimately giving priority to protecting high-value corridors. Circuit flow is combined with multi-path optimization, and dynamic environments and time series are introduced to achieve intelligent adjustments in time and space. At the same time, deep learning is used to optimize the objective function to balance resistance, gradient, and dynamic factors.

[0094] Step S4: Dynamic identification of ecological pinch points and obstacle points.

[0095] Based on the ecological resistance surface constructed in step S2, identify pinch points and obstacle points in the ecosystem, i.e., bottlenecks and blocking areas in the migration path;

[0096] In step S4, based on the circuit theory model, the current density in different areas is calculated to identify high-importance nodes in the ecological network, i.e., pinch points, and obstacle points that hinder species migration;

[0097] In step S4, the positions and priorities of the pinch points and the obstacle points are adjusted according to the changes in mining, including the expansion of the mining area and the destruction of vegetation;

[0098] In step S4, the vegetation density changes and meteorological changes collected in step S1 are combined, and the agent-based simulation method is used to simulate the frequency and pressure of different species passing through the pinch points and obstacle points in the channel, and the positions of the pinch points and obstacle points are dynamically adjusted;

[0099] In step S4, the current density is calculated based on the resistance surface:

[0100] Using circuit theory, the current distribution is simulated on the ecological resistance surface and the current density in each area is calculated:

[0101] Where J(x, y) represents the current density of pixel (x, y), that is, the pressure level of the point as a migration path in the ecological network, S, T represent the starting point set S and the end point set T, respectively defining the source and target area of ​​migration, V s , V t represents the voltage value at the starting point s and the end point t, G(x, y) represents the conduction rate of the pixel (x, y), which is defined as Inversely proportional to the resistance value, represents the resistance value of pixel (x, y) on the ecological resistance surface. Here, the current density J(x, y) reflects the importance of the connectivity of pixel (x, y) in the migration network in the ecosystem. A high current density indicates a bottleneck position in the path that is critical to migration.

[0102] Based on the current density map, identify the current high value concentration area as the pinch point: C(x, y) = I(J(x, y)>θ J), where C(x, y) indicates whether the pixel (x, y) is a pinch point, 1 indicates a pinch point, 0 indicates a non-pinch point, I(·) indicates an indicator function, outputs 1 when the condition is met, otherwise outputs 0, θ J represents the current density threshold, which is used to define the importance of the pinch point. The area with current density higher than the threshold is marked as the pinch point, which represents the key bottleneck point in the migration path;

[0103] Conductivity and resistance change rate are used to detect obstacle points, i.e., blocked areas:

[0104] Among them, B(x, y) indicates whether the pixel (x, y) is an obstacle point, 1 indicates an obstacle point, and 0 indicates a non-obstacle point. Indicates the resistance value change rate of pixel (x, y), indicating the dynamic change trend of the degree of obstruction, θ B It represents the resistance change rate threshold, which is used to define the criteria for determining the obstacle point. The area where the resistance value change rate is higher than the threshold is marked as an obstacle point, representing a dynamically changing area that may hinder the migration of species.

[0105] In step S4, the pinch points and obstacle points are adjusted dynamically according to the mining conditions:

[0106] According to the mining expansion area, the current density threshold is redefined: θ' J (t) = θ J ·(1+α·A mine (t)), where θ' J (t) represents the pinch current density threshold after dynamic adjustment at time t, α represents the influence coefficient of mine expansion on current density, A mine (t) represents the proportion of the mine expansion area at time t. As the mine expansion area increases, the current density threshold is increased to avoid misjudging non-critical nodes as pinch points;

[0107] According to the dynamic changes of vegetation destruction, the obstacle point judgment condition is adjusted, θ' B (t) = θ B ·(1+β·L veg (t)), where θ' B (t) represents the resistance change rate threshold after dynamic adjustment at time t, β represents the influence coefficient of vegetation destruction on the resistance change rate, L veg (t) represents the vegetation coverage loss rate at time t. The higher the vegetation destruction rate, the lower the judgment threshold and the more sensitive the identification of possible obstacles.

[0108] In step S4, the pinch points and obstacle points are dynamically adjusted based on the agent simulation:

[0109] Agent-Based Modeling is used to simulate the frequency of species passing through pinch points in migration corridors:

[0110] Among them, F C (t) represents the frequency of agents passing through the pinch point C at time t, N A represents the total number of agents participating in the migration in the simulation, A i (t) represents the position of agent i at time t, I(A i (t)∈C) means that if agent i passes through pinch point C, output 1, otherwise output 0. The frequency of the pinch point reflects the importance of the point in the migration network. The pinch points with high frequency need to be protected first;

[0111] Simulate the stress level of the agent passing through the obstacle point: Among them, P B (t) represents the agent pressure level at obstacle point B at time t, It represents the resistance value of the location of agent i. The pressure level of the obstacle point reflects its negative impact on the migration path. The obstacle points with high pressure need to be repaired or alleviated first;

[0112] Specifically, in step S4, current density and resistance change rate are used to identify pinch points and obstacle points, and the identification threshold is dynamically adjusted to adapt to changes in mine expansion and vegetation destruction. At the same time, the importance of pinch points and the pressure level of obstacle points are quantified in combination with agent simulation, and key locations and repair strategies are dynamically adjusted in priority;

[0113] The ecological corridor in this article refers to a channel connecting two or more habitats in an ecological network, which can provide guarantees for species migration, population exchanges and functional flow of ecosystems.

[0114] Step S5: Ecological management based on dynamic results.

[0115] Based on the ecological corridors, pinch points and obstacle points identified in step S3 and step S4, targeted ecological protection and restoration measures are taken; the identification results are transmitted to the ecological management system in real time, and the mine management can give priority to repairing or protecting key corridors and pinch points to reduce the impact of human activities on ecological connectivity.

[0116] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A mine ecological safety identification method based on a convolutional neural network model, characterized in that: include, Step S1, mine data collection, Collect mine data, including topographic data, vegetation density, meteorological data, and mining record data; Step S2, dynamic resistance surface modeling, Based on the mine data collected in step S1, a dynamic ecological resistance surface is constructed to reflect the spatial distribution of the degree of obstruction of species migration in each region of the ecosystem, and the resistance value of each region depends on the ecological factors; Step S3, multi-path ecological corridor identification, By combining deep learning with circuit theory, multiple ecological corridors, i.e. possible migration paths, are identified and generated on the ecological resistance surface; In step S3, based on the deep learning model combined with time series analysis, the priority of each path is adjusted to give priority to protecting corridors with high ecological value and high-demand species passages; Step S4: Dynamic identification of ecological pinch points and obstacle points. Based on the ecological resistance surface constructed in step S2, identify pinch points and obstacle points in the ecosystem, i.e., bottlenecks and blocking areas in the migration path; Step S5: Ecological management based on dynamic results. Based on the ecological corridors, pinch points and obstacle points identified in step S3 and step S4, targeted ecological protection and restoration measures are taken.

2. According to claim 1, a mine ecological safety identification method based on a convolutional neural network model is characterized in that: In step S2, a convolutional neural network (CNN) is used to dynamically adjust the resistance value according to the terrain data, vegetation density, and water distribution in the meteorological data.

3. According to claim 2, a mine ecological safety identification method based on a convolutional neural network model is characterized in that: In step S2, based on the data collected in step S1, a convolutional neural network (CNN) is used to dynamically adjust the resistance value. The specific steps include: Rasterize terrain data spatially: R raw (x, y) = f norm (T(x,y))+g norm (V(x,y))+h norm (W(x, y)), where R raw (x, y) represents the initial resistance value of the area (x, y), f norm (·), g norm (·), h norm (·) represents the function for normalizing each ecological factor, keeping the value range between [0, 1]. T(x, y) represents the terrain slope value of the region (x, y). The greater the slope, the higher the resistance. V(x, y) represents the vegetation coverage density of the region (x, y). The higher the coverage, the lower the resistance. W(x, y) represents the water distribution characteristics of the region (x, y). The closer to the water body, the lower the resistance. Based on the initial resistance value R raw (x, y), and dynamically adjust it using a convolutional neural network (CNN), capturing local spatial characteristics and reflecting the interactive relationship of ecological factors, and dynamically adjusting resistance: Among them, R dyn (x, y) represents the dynamically adjusted resistance value of the region (x, y), σ(·) represents the ReLU activation function, which is used to maintain the nonlinear characteristics of the output, and K i，j Represents the weight matrix of the convolution kernel, representing the weights of different positions, k represents the size of the convolution kernel, b represents the bias term, which is used to dynamically adjust the baseline of the overall resistance value, R raw (x+i, y+j) represents the local neighborhood pixel value of the initial resistance value.

4. According to claim 3, a mine ecological safety identification method based on a convolutional neural network model is characterized in that: In step S2, an environmental change response mechanism is introduced: Introducing the time dimension t and the dynamic factor D(x, y, t), capturing the impact of environmental changes based on the time series, and calculating the dynamic resistance value of the time dimension: R final (x, y, t) = R dyn (x, y)·(1+α·D(x, y, t)), where, R final (x, y, t) represents the final dynamic resistance value of the area (x, y) at time t, α represents the weight coefficient of the dynamic factor, which is used to adjust the influence of environmental changes on the resistance value, and D(x, y, t) represents the dynamic environmental factor of the area (x, y) at time t, which is obtained from the measured data and ranges from [-1, 1]. The resistance surface diagram is generated: in, represents the dynamic ecological resistance surface at time t, Ω represents the research scope of the ecosystem, i.e., the total area, and R final (x, y, t) represents the final resistance value for each pixel; the resulting spatial distribution map Dynamically reflect the migration resistance in the ecological environment.

5. According to claim 4, a mine ecological safety identification method based on a convolutional neural network model is characterized in that: In step S3, multi-path identification of ecological corridors is performed. In the dynamic ecological resistance On the other hand, circuit theory is used to calculate the flow of current when species migrate. Among them, I(x, y) represents the current intensity of pixel (x, y), which represents the flow intensity in the ecological migration network. represents the dynamic ecological resistance value generated in step S2, S and T represent the migration starting point set and target point set in the ecosystem, V s , V t Represents the voltage values ​​of the starting point s and the target point t, which are set to 1 and 0; Using the current flow model I(x, y), potential migration paths are identified: P k (x, y) = H (I (x, y) - θ k ), P k (x, y) represents the indication value of path k at pixel (x, y), 1 means the path passes through the pixel, 0 means it does not pass through, H(·) represents the unit step function, when the input is greater than 0, the output is 1, otherwise it outputs 0, θ k Represents the current threshold of path k, which is used to control the path flow, and the control threshold θ k Adjust the number of paths to generate multiple evenly distributed migration paths; introduce a deep learning model to dynamically optimize path weights based on ecological factors: in, represents the optimization objective function, which is the total resistance and negative impact of all paths, N represents the total number of candidate migration paths, λ1, λ2, λ3 represent weight parameters, which measure the influence of resistance, terrain gradient and dynamic factors respectively. represents the terrain gradient of pixel (x, y), corresponding to the slope cost of migration, and D(x, y, t) represents the dynamic environmental factor of pixel (x, y) at time t.

6. The method for identifying mine ecological safety based on a convolutional neural network model according to claim 5 is characterized in that: In step S3, the path priority is adjusted based on the time series: Combine the time series model to predict the importance of each path and dynamically adjust the priority: Among them, W k (t) represents the priority of path k at time t, β1, β2 represent the priority adjustment weights, and H k (t) represents the dynamic demand index of path k, the species migration demand based on time series prediction, and dxdy represents the integration unit in two-dimensional space; Use time series analysis to predict future migration demand: H k (t+1) = γ·H k (t)+δ·Env(t), where H k (t+1) represents the migration demand of path k at time t+1, γ represents the time attenuation factor of the migration demand, δ represents the response coefficient of environmental change, Env(t) represents the real-time change of environmental factors at time t, Combined with the path priority, select high ecological value paths for protection: P prior = argmax k W k (t), where P prior represents the set of paths to be protected first, W k (t) represents the current priority of path k, dynamically selecting migration paths with high ecological value and high demand for priority protection and restoration.

7. The method for identifying mine ecological safety based on a convolutional neural network model according to claim 6 is characterized in that: In step S4, based on the circuit theory model, the current density in different areas is calculated to identify high-importance nodes in the ecological network, i.e., pinch points, and obstacles that hinder species migration.

8. The method for identifying mine ecological safety based on a convolutional neural network model according to claim 7 is characterized in that: In step S4, the positions and priorities of the pinch points and the obstacle points are adjusted according to the changes in mining, including the expansion of the mining area and the destruction of vegetation; In step S4, the vegetation density changes and meteorological changes collected in step S1 are combined, and the agent-based simulation method is used to simulate the frequency and pressure of different species passing through pinch points and obstacle points in the channel, and the positions of pinch points and obstacle points are dynamically adjusted.

9. The method for identifying mine ecological safety based on a convolutional neural network model according to claim 8 is characterized in that: In step S4, the current density is calculated based on the resistance surface: Using circuit theory, the current distribution is simulated on the ecological resistance surface and the current density in each area is calculated: Where J(x, y) represents the current density of pixel (x, y), that is, the pressure level of the point as a migration path in the ecological network, S, T represent the starting point set S and the end point set T, respectively defining the source and target area of ​​migration, V s , V t represents the voltage value at the starting point s and the end point t, G(x, y) represents the conduction rate of the pixel (x, y), which is defined as Inversely proportional to the resistance value, represents the resistance value of pixel (x, y) on the ecological resistance surface, where the current density J(x, y) reflects the importance of the connectivity of pixel (x, y) in the migration network of the ecosystem; Based on the current density map, identify the current high value concentration area as the pinch point: C(x, y) = I(J(x, y)>θ J ), where C(x, y) indicates whether the pixel (x, y) is a pinch point, 1 indicates a pinch point, 0 indicates a non-pinch point, I(·) indicates an indicator function, outputs 1 when the condition is met, otherwise outputs 0, θ J represents the current density threshold, which is used to define the importance of the pinch point. The area with current density higher than the threshold is marked as the pinch point, which represents the key bottleneck point in the migration path; Conductivity and resistance change rate are used to detect obstacle points, i.e., blocked areas: Among them, B(x, y) indicates whether the pixel (x, y) is an obstacle point, 1 indicates an obstacle point, and 0 indicates a non-obstacle point. Indicates the resistance value change rate of pixel (x, y), indicating the dynamic change trend of the degree of obstruction, θ B It represents the resistance change rate threshold, which is used to define the criteria for determining obstacle points. Areas where the resistance value change rate is higher than the threshold are marked as obstacle points, representing dynamically changing areas that may hinder species migration.

10. A mine ecological safety identification method based on a convolutional neural network model according to claim 9, characterized in that: In step S4, the pinch points and obstacle points are adjusted dynamically according to the mining conditions: According to the mining expansion area, the current density threshold is redefined: θ' J(t) =θ J ·(1+α·A mine (t)), where θ' J (t) represents the pinch current density threshold after dynamic adjustment at time t, α represents the influence coefficient of mine expansion on current density, A mine (t) represents the proportion of the mine expansion area at time t, as the mine expansion area increases; According to the dynamic changes of vegetation destruction, the obstacle point judgment condition is adjusted, θ' B (t) = θ B ·(1+β·L veg (t)), where θ' B (t) represents the resistance change rate threshold after dynamic adjustment at time t, β represents the influence coefficient of vegetation destruction on the resistance change rate, L veg (t) represents the vegetation cover loss rate at time t. The higher the vegetation destruction rate, the lower the judgment threshold; In step S4, the pinch points and obstacle points are dynamically adjusted based on the agent simulation: Agent-Based Modeling is used to simulate the frequency of species passing through pinch points in migration corridors: Among them, F C (t) represents the frequency of agents passing through the pinch point C at time t, N A represents the total number of agents participating in the migration in the simulation, A i (t) represents the position of agent i at time t, I(A i (t)∈C) means that if agent i passes through pinch point C, output 1, otherwise output 0. The frequency of the pinch point reflects the importance of the point in the migration network. The pinch points with high frequency need to be protected first; Simulate the stress level of the agent passing through the obstacle point: Among them, P B (t) represents the agent pressure level at obstacle point B at time t, It represents the resistance value at the location of agent i. The pressure level of the obstacle point reflects its negative impact on the migration path. Obstacle points with high pressure need to be repaired or alleviated first.

Images