Ecological management evaluation method based on ecological pattern and ecological integration monitoring system

Abstract

The present application relates to the field of ecological protection data processing, in particular to an ecological management evaluation method based on ecological pattern and an ecological integrated monitoring system. The method proposed by the present application determines an ecological pattern evaluation matrix based on six dimensions according to an ecological pattern data set, and obtains a dynamic weight distribution model based on an ecological management target. The weight matrix obtained in the dynamic weight distribution model is multiplied by the ecological pattern evaluation matrix to obtain the evaluation score based on different ecological management targets, so that the completion of the ecological management target can be judged. The method of the present application focuses on the ecological pattern level, comprehensively integrates multi-dimensional ecological information, and systematically reflects the overall characteristics and operating state of the ecological system. A unified weight adjustment model is used to flexibly adapt to different management targets such as ecological protection, ecological planning and ecological restoration, and there is no need to build a complex evaluation system for each target.

Description

Technical Field

[0001] This invention relates to the field of ecological protection data processing, and more specifically, to an ecological management assessment method based on ecological patterns and an integrated ecological monitoring system. Background Technology

[0002] In recent years, the importance of ecological management has become increasingly prominent. With the rapid development of human society, the ecological environment faces many severe challenges, such as deforestation, wetland degradation, and biodiversity loss. The ecosystem is the foundation upon which life on Earth depends. It not only provides humans with basic survival resources such as food, water, and air, but also maintains important ecological service functions such as climate regulation, soil and water conservation, and soil fertility maintenance.

[0003] Several problems urgently need to be addressed in the assessment of ecological management. First, there is a lack of unified and effective quantitative indicators. Currently, ecological management assessments based on different objectives involve numerous indicators, but these indicators are often difficult to compare and measure uniformly across different regions and ecosystems. Second, manual statistical analysis has many drawbacks. Manual collection and processing of ecological data is prone to subjective bias and recording errors, and manual statistical analysis is inefficient, failing to reflect rapid changes in ecosystems in a timely manner, resulting in delayed assessment results that cannot meet the timeliness requirements of ecological management. These problems severely restrict the accuracy and effectiveness of ecological management. Therefore, a method is needed that can effectively assess the effectiveness of ecological management under different ecological management objectives. Summary of the Invention

[0004] The purpose of this invention is to provide an ecological management assessment method and an integrated ecological monitoring system based on ecological patterns. The method proposed in this invention can effectively assess the effectiveness of ecological management work under different ecological management objectives.

[0005] According to a first aspect of the present invention, an ecological management assessment method based on ecological patterns is proposed, comprising:

[0006] S1. A dynamic weight allocation model is obtained by training on an ecological management goal and management operation dataset;

[0007] S2. Determine the first weight matrix based on the dynamic weight allocation model. ;

[0008] S3. Determine the ecological pattern evaluation matrix based on the ecological pattern dataset of the target area. ;

[0009] S4. Determine the ecological pattern assessment score SC = W1 × D1 based on the first weight matrix W1 and the ecological pattern evaluation matrix D1;

[0010] S5. Determine the level of achievement of ecological management objectives based on the ecological pattern assessment score (SC).

[0011] According to some embodiments, in the method of the first aspect of the present invention, the ecological management objective includes any one of ecological planning T1, ecological protection T2, and ecological restoration T3.

[0012] Dynamic weight allocation models include reinforcement learning models, which include state space and action space.

[0013] The state space records ecosystem parameters, including the ecological connectivity ratio, ecosystem composition ratio, ecosystem type area change rate, ecosystem spatial pattern characteristics, ecosystem comprehensive dynamic degree, ecosystem classification dynamic degree, as well as ecological planning T1, ecological protection T2, and ecological restoration T3; it is used to enable agents to perceive the current environmental status of the ecosystem and to provide a basis for agent decision-making.

[0014] The action space includes a first weight matrix W1, and the value range and adjustment step size of each data in the first weight matrix W1, which are used to limit the action mode of the agent so that the agent knows what actions are available.

[0015] According to some embodiments, in the method of the first aspect of the present invention, step S1. training based on the management operation dataset to obtain a dynamic weight allocation model includes:

[0016] Different reward functions should be set for different ecological management objectives. ;

[0017] Preprocess the management operation dataset to obtain training samples;

[0018] The agent in the reinforcement learning model selects a first weight matrix W1 from the action space as the action to be executed based on the current state;

[0019] The impact of the selected first weight matrix W1 on the ecological management objective is evaluated according to the reward function, and the reward signal and new state information are fed back to the agent.

[0020] The agent updates and learns its weight allocation plan using reinforcement learning algorithms based on reward signals and new state information to determine the final reinforcement learning model.

[0021] According to some embodiments, in the method of the first aspect of the present invention, the agent of the reinforcement learning model is the core component for performing decision-making and learning, and is used to select appropriate actions from its action space to perform by perceiving the current state of the environment, and then adjust its behavioral strategy according to the reward signal fed back by the environment.

[0022] According to some embodiments, in the method of the first aspect of the present invention, the state space further includes a connectivity index COH, a first aggregation index AI, a total number of patches NP, an average patch area, a boundary density ED, and a second aggregation index C; the action space further includes a second weight matrix and a third weight matrix; and the method of the first aspect of the present invention further includes:

[0023] The second weight matrix W2 is determined based on the dynamic weight allocation model. and the third weight matrix W3= .

[0024] According to some embodiments, in the method of the first aspect of the present invention, step S3, determining the ecological pattern evaluation matrix D1 based on the ecological pattern dataset of the target area, includes:

[0025] Based on the ecological pattern dataset, the connectivity index (COH), the first clustering index (AI), the total number of patches (NP), and the average patch area were calculated. Boundary density ED, second aggregation index C;

[0026] The ecological connectivity ratio is calculated based on the connectivity index (COH) and the first aggregation index (AI), following the formula: ;

[0027] Based on the total number of plaques (NP) and the average plaque area The spatial pattern characteristics of the ecosystem are calculated using the boundary density ED and the second aggregation index C, following the formula: .

[0028] According to some embodiments, in the method of the first aspect of the present invention, step S5. determining the completion level of the ecological management objectives based on the ecological pattern assessment score SC includes:

[0029] If the ecological pattern assessment score SC is less than or equal to the first score, the completion level is poor.

[0030] If the ecological pattern assessment score SC is greater than the first score and less than or equal to the second score, the completion level is average.

[0031] If the ecological pattern assessment score (SC) is greater than the second score and less than or equal to the third score, the completion level is good.

[0032] If the ecological pattern assessment score (SC) is greater than the third score, the completion level is rated as excellent.

[0033] According to a second aspect of the present invention, an ecological management assessment device based on ecological patterns is provided, comprising:

[0034] The evaluation matrix generation module is used to determine the ecological pattern evaluation matrix D1 based on the ecological pattern dataset of the target area. ;

[0035] in, Indicates ecological connectivity. Indicates the proportion of ecosystem components. Indicates the rate of change in the area of ​​ecosystem types. Indicates the spatial pattern characteristics of the ecosystem. Indicates the overall dynamic degree of the ecosystem. Indicates the dynamic degree of ecosystem classification;

[0036] The weight model training module is used to train a dynamic weight allocation model based on the management operation dataset.

[0037] The weight matrix generation module is used to determine the first weight matrix based on the dynamic weight allocation model. ;

[0038] The scoring module is used to determine the ecological pattern assessment score SC=W1 × D1 based on the first weight matrix W1 and the ecological pattern evaluation matrix D1.

[0039] The rating module is used to determine the level of achievement of ecological management objectives based on the ecological pattern assessment score (SC).

[0040] According to a third aspect of the present invention, an integrated ecological monitoring system is proposed, comprising:

[0041] The data acquisition unit includes a satellite remote sensing monitoring module, a drone aerial photography monitoring module, and an Internet of Things acquisition module, which are used to collect ecological pattern datasets and management operation datasets.

[0042] The data analysis unit includes an ecological management assessment device based on ecological pattern as described in the second aspect of the present invention, for performing an ecological management assessment method based on ecological pattern as described in the first aspect of the present invention.

[0043] The data visualization unit includes a data visualization platform used to display the analysis results of the data analysis unit.

[0044] The solution proposed in this invention has the following beneficial effects:

[0045] 1. The method proposed in this invention employs a unified weight adjustment model, flexibly adapting to different management objectives such as ecological protection, ecological planning, and ecological restoration, without requiring the construction of a complex assessment system for each objective. By dynamically adjusting weights, it can accurately highlight key ecological elements under different management objectives, making the assessment process more targeted and effective. This strong adaptability improves the efficiency of assessment work, saves resources, and ensures the reliability and consistency of assessment results under changing ecological management scenarios, providing strong support for ecological management decision-making.

[0046] 2. The method proposed in this invention focuses on the ecological pattern level, comprehensively integrating multi-dimensional ecological information to systematically reflect the overall characteristics and operational status of the ecosystem. From the perspective of ecological pattern, it can comprehensively analyze the interrelationships and spatial distribution patterns of various elements of the ecosystem, avoiding the one-sidedness of single-indicator assessment. It provides a more macroscopic and comprehensive perspective for ecological management, which helps to formulate more systematic and scientific ecological management strategies and achieve comprehensive protection and sustainable development of the ecosystem.

[0047] 3. The method proposed in this invention can quickly calculate the evaluation score based on real-time acquired ecological pattern data, and reflect the effectiveness of ecological management work in real time through dynamic weight allocation, promptly identifying problems and shortcomings in the management process. Compared with traditional evaluation methods, its timeliness advantage is obvious, providing a scientific basis for timely adjustment of ecological management strategies and effectively improving the accuracy and flexibility of management work. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used 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 those skilled in the art, other drawings can be obtained based on these drawings without exceeding the scope of protection claimed by the present invention.

[0049] Figure 1 This is a flowchart illustrating an ecological management assessment method 1000 based on ecological patterns according to the present invention.

[0050] Figure 2 This is a flowchart illustrating an ecological management assessment method 2000 based on ecological patterns according to the present invention.

[0051] Figure 3 for Figure 1 A flowchart illustrating the detailed steps of step S1 in an ecological management assessment method 1000 based on ecological patterns.

[0052] Figure 4 for Figure 1A flowchart illustrating the detailed steps of step S3 in an ecological management assessment method 1000 based on ecological patterns.

[0053] Figure 5 for Figure 1 A flowchart illustrating the detailed steps of step S5 in an ecological management assessment method 1000 based on ecological patterns.

[0054] Figure 6 This is a schematic diagram of the structure of an ecological management assessment device 3000 based on ecological pattern according to the present invention;

[0055] Figure 7 This is a schematic diagram of the structure of an integrated ecological monitoring system 4000 according to the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Figure 1 This is a flowchart illustrating an ecological management assessment method 1000 based on ecological patterns according to the present invention. Figure 1 As shown, method 1000 includes steps S1-S5.

[0058] In step S1, the ecological management assessment device based on the ecological pattern (e.g., a processor, which will be used as an example below) is trained based on the ecological management objectives and management operation dataset to obtain a dynamic weight allocation model.

[0059] In some specific embodiments, in step S1, the dynamic weight allocation model is a reinforcement learning model, which includes a state space and an action space.

[0060] In some specific embodiments, in step S1, the state space of the reinforcement learning model is defined. Optionally, the state space includes ecological connectivity ratios, ecosystem composition ratios, ecosystem type area change rates, ecosystem spatial pattern characteristics, ecosystem comprehensive dynamics, ecosystem classification dynamics, and ecological planning T1, ecological protection T2, and ecological restoration T3. Optionally, the state space also includes other environmental variables related to ecological management, such as regional climate conditions and the intensity of human activities, to more comprehensively reflect the actual state of the ecosystem and external influencing factors.

[0061] In some specific embodiments, in step S1, the action space of the reinforcement learning model is defined. The action space represents the weight allocation strategy for the six dimensions of data in the ecological pattern evaluation matrix. Optionally, the action space includes a first weight matrix W1, and the value range and adjustment step size of each data in the first weight matrix W1. For example, the value range of the weight data can be set in the interval [0,1], and the adjustment step size for iteratively updating each weight data is 0.01. By reasonably setting the value range and adjustment step size of the weight data, the flexibility and accuracy of weight allocation are ensured.

[0062] In some specific embodiments, in step S1, the processor sets different reward functions for different ecological management objectives. .

[0063] In some specific embodiments, in step S1, the data in the management operation dataset is preprocessed, including data cleaning and data normalization, to obtain training samples. Optionally, the management operation dataset includes historical ecological data under different ecological management objectives and corresponding management operation records, specifically including the changes in the six dimensions of the ecological pattern evaluation matrix at different time points, and the response results under different management operations (such as the implementation of ecological restoration projects, the adjustment of ecological protection policies, etc.).

[0064] In some specific embodiments, in step S1: the processor preprocesses the management operation dataset to obtain training samples; the agent of the reinforcement learning model selects a first weight matrix W1 from the action space as an action to be executed according to the current state; the processor evaluates the impact of the selected first weight matrix W1 on the ecological management goal according to the reward function, and feeds back the reward signal and new state information to the agent; the agent updates and learns its weight allocation plan using the reinforcement learning algorithm according to the reward signal and new state information, so as to determine the final reinforcement learning model.

[0065] In some specific embodiments, in step S1, the dynamic weight allocation model is a linear regression model. In some specific embodiments, it is assumed that there is a linear relationship between the completion status of ecological management goals and the six dimensions of data in the ecological pattern evaluation matrix; that is, the goal completion score can be represented as a linear combination of the six dimensions. Using a linear regression algorithm, historical data from the management operation dataset is used, with the six dimensions as independent variables and the actual completion status of ecological management goals as the dependent variable, to train the model and obtain the regression coefficients corresponding to each dimension. These regression coefficients can then serve as the mapping matrix between ecological management goals and weight allocation.

[0066] In some specific embodiments, in step S1, the dynamic weight allocation model is an artificial neural network model. In some specific embodiments, the complex mapping relationship between the six dimensions of data and the completion status of ecological management goals is learned through training. During the training process, the connection weights of the neural network are continuously adjusted, eventually converging to weight values ​​that can better predict the completion status of the goals. These weight values ​​can reflect the importance of the six dimensions of data in the ecological pattern evaluation matrix.

[0067] In step S2, the first weight matrix is ​​determined according to the dynamic weight allocation model. In some specific embodiments, in step S2, the processor can obtain the specific value of the first weight matrix W1 corresponding to the ecological management objective based on the already trained reinforcement learning model. For example, in some specific embodiments, the ecological management objective is ecological planning T1, and the first weight matrix... The ecological management objective is ecological protection T2, with the first weight matrix. The ecological management objective is ecological restoration T3, with the first weight matrix. .

[0068] In step S3, the processor determines the ecological pattern evaluation matrix based on the ecological pattern dataset of the target area. .

[0069] in, Indicates the ratio of ecological connectivity. Indicates the proportion of ecosystem components. Indicates the rate of change in the area of ​​ecosystem types. Indicates the spatial pattern characteristics of the ecosystem. Indicates the overall dynamic degree of the ecosystem. This indicates the dynamic degree of ecosystem classification.

[0070] In some specific embodiments, the processor calculates the connectivity index (COH) and the first clustering index (AI) based on the ecological pattern dataset, and further calculates the ecological connectivity ratio based on the connectivity index (COH) and the first clustering index (AI). .

[0071] In some specific embodiments, the processor sets the second weight matrix W2 by a preset method. In some specific embodiments, the processor obtains the second weight matrix W2 based on a dynamic weight allocation model.

[0072] In some specific embodiments, the processor calculates the ecological connectivity ratio from the connectivity index (COH) and the first aggregation index (AI) according to the following formula:

[0073] .

[0074] In some specific embodiments, the connectivity index (COH) primarily reflects the connectivity between different ecological units within an ecosystem, such as the degree of physical connection between habitat patches, while the first aggregation index (AI) focuses on reflecting the spatial aggregation of ecological elements. By combining the two to calculate the ecological connectivity ratio, the connectivity status of an ecosystem can be assessed more comprehensively and accurately. For example, an ecosystem may have a high connectivity, indicating numerous connection channels between habitat patches. However, if these patches have low aggregation and are scattered, the overall connectivity of the ecosystem may not be ideal. The ecological connectivity ratio can comprehensively reflect this complex situation.

[0075] In some specific embodiments, the processor calculates the ecosystem composition ratio based on the ecological pattern dataset, following the formula: ;

[0076] in, TS represents the total area of ​​the target area being assessed, where i represents the area of ​​ecosystem type i.

[0077] In some specific embodiments, the processor calculates the rate of change of ecosystem area based on the ecological pattern dataset, following the formula: ;

[0078] in, This represents the initial area of ​​ecosystem type i. This represents the area of ​​ecosystem type i at the end of the period.

[0079] In some specific embodiments, the processor calculates the total number of patches (NP) and the average patch area based on the ecological pattern dataset. The boundary density ED and the second aggregation index C are used to further calculate the spatial pattern characteristics of the ecosystem.

[0080] In some specific embodiments, the processor sets the third weight matrix in a preset manner. In some specific embodiments, the processor obtains the third weight matrix based on a dynamic weight allocation model. .

[0081] In some specific embodiments, the processor bases the total number of plaques (NP) and the average plaque area on... The spatial pattern characteristics of the ecosystem are calculated using the boundary density ED and the second aggregation index C, following the formula: .

[0082] In some specific embodiments, the processor calculates the integrated dynamic degree of the ecosystem based on the ecological pattern dataset, following the formula: ;

[0083] Wherein, ECOi represents the initial area of ​​ecosystem type i. This represents the absolute value of the area where ecosystem type i is converted to a non-i type ecosystem.

[0084] In some specific embodiments, the processor calculates the ecosystem classification dynamics based on the ecological pattern dataset, following the formula: ;

[0085] Where i represents the initial ecosystem type and j represents the final ecosystem type. Let Aij represent the area of ​​each ecosystem type, Aij represent the proportion of the i-th ecosystem type at the beginning of the period that transformed into the j-th ecosystem type at the end of the period, and Biij represent the proportion of the j-th ecosystem type at the end of the period that transformed from the i-th ecosystem type at the beginning of the period.

[0086] In step S4, the ecological pattern assessment score SC = W1 × D1 is determined based on the first weight matrix W1 and the ecological pattern evaluation matrix D1. In some specific embodiments, the ecological pattern assessment score SC = .

[0087] In step S5, the completion level of the ecological management objectives is determined based on the ecological pattern assessment score SC. In some specific embodiments, if the ecological pattern assessment score SC is less than or equal to the first score, the completion level is poor; if the ecological pattern assessment score SC is greater than the first score and less than or equal to the second score, the completion level is average; if the ecological pattern assessment score SC is greater than the second score and less than or equal to the third score, the completion level is good; and if the ecological pattern assessment score SC is greater than the third score, the completion level is excellent.

[0088] Optionally, the first, second, and third rating values ​​are 0.5, 0.6, and 0.8, respectively.

[0089] According to such Figure 1 The implementation method shown in this invention uses a unified weight adjustment model to flexibly adapt to different management objectives such as ecological protection, ecological planning, and ecological restoration, without the need to build a complex evaluation system for each objective. It focuses on the ecological pattern level, comprehensively integrates multi-dimensional ecological information, and systematically reflects the overall characteristics and operating status of the ecosystem. It quickly calculates the evaluation score based on the real-time acquired ecological pattern data, and reflects the effectiveness of ecological management work in real time through dynamic weight allocation, so as to promptly identify problems and deficiencies in the management process.

[0090] Figure 2This is a flowchart illustrating an ecological management assessment method 2000 based on ecological patterns according to the present invention. Figure 2 As shown, method 2000 includes steps S201-S206. Among them, steps S202 and steps S205-S206 are the same as steps S2 and steps S4-S5 in method 1000, and will not be described again here.

[0091] In step S201, a dynamic weight allocation model is obtained by training based on the ecological management goals and management operation dataset.

[0092] In some specific embodiments, in step S201, the dynamic weight allocation model is a reinforcement learning model, and the state space of the reinforcement learning model further includes the connectivity index COH, the first clustering index AI, the total number of patches NP, and the average patch area. The boundary density ED, the second clustering index C, and the action space also include the second weight matrix. and the third weight matrix .

[0093] In some specific embodiments, in step S201, reward functions are designed under different ecological management objectives, wherein the reward function includes a target reward function. Ecological connectivity ratio reward function and reward function of spatial pattern characteristics of ecosystem .

[0094] Optionally, the target reward function Different combinations can be configured according to specific ecological management objectives; see [link / reference]. Figure 4 Step S22. Optionally, the ecological connectivity ratio reward function. = Ecosystem spatial pattern characteristics reward function = .in, , , , , , These are the connectivity index (COH), the first aggregation index (AI), the total number of patches (NP), and the average patch area. The changes in boundary density ED and the second aggregation index C.

[0095] In some specific embodiments, in step S201, the agent of the reinforcement learning model selects a combination of a first weight matrix W1, a second weight matrix W2, and a third weight matrix W3 from the action space as an action to be executed based on the current state. In some specific embodiments, the reinforcement learning model includes a decision-making subject, i.e., an agent, which learns by interacting with the environment. In some specific embodiments, the agent in the reinforcement learning model is placed in a simulated ecological management environment, and the agent selects a weight allocation strategy (i.e., a combination of the first weight matrix W1, the second weight matrix W2, and the third weight matrix W3) from the action space as an action to be executed based on the current state (including six-dimensional data, ecological management goals, etc.).

[0096] In some specific embodiments, in step S201, the impact of the selected weight allocation strategy on the ecological management objective is evaluated according to the reward function, and the reward signal and new state information are fed back to the agent. In some specific embodiments, the reward signal is the change of the six dimensions of data in the ecological pattern evaluation matrix after weight adjustment, and the state information is the corresponding response of the ecosystem. In step S201, the agent updates and learns its weight allocation plan using a reinforcement learning algorithm based on the reward signal and the new state information to determine the final reinforcement learning model. In some specific embodiments, the reinforcement learning algorithm includes any one of the following: Q-learning algorithm, deep deterministic policy gradient algorithm, SARSA algorithm, and deep Q network.

[0097] In step S203, the second weight matrix is ​​determined according to the dynamic weight allocation model. and the third weight matrix

[0098] In some specific embodiments, the processor can obtain the second weight matrix W2 and the third weight matrix corresponding to the ecological management objective based on the already trained reinforcement learning model. The specific value.

[0099] In step S204, the ecological pattern evaluation matrix is ​​determined based on the ecological pattern dataset of the target area. .

[0100] in, Indicates the ratio of ecological connectivity. Indicates the proportion of ecosystem components. Indicates the rate of change in the area of ​​ecosystem types. Indicates the spatial pattern characteristics of the ecosystem. Indicates the overall dynamic degree of the ecosystem. This indicates the dynamic degree of ecosystem classification.

[0101] In some specific embodiments, in step S204, the calculation of the ecological connectivity ratio based on the ecological pattern dataset follows the formula: .

[0102] When assessing the ecological condition of a region, the ecological connectivity ratio can help identify bottlenecks in connectivity among various ecosystems within the region. For example, some areas may have relatively high connectivity, but due to low aggregation, ecological flows (such as species migration and gene flow) are limited, becoming a key limiting factor for ecosystem connectivity. Conversely, some areas may have high aggregation, but insufficient connectivity can also affect ecosystem connectivity. By analyzing the ecological connectivity ratio, these key areas can be precisely located, providing clear priorities and directions for ecological restoration and protection efforts.

[0103] In some specific embodiments, in step S204, the spatial pattern characteristics of the ecosystem are calculated based on the ecological pattern dataset, following the formula: .

[0104] When assessing the ecological condition of a region: the total number of patches (NP) reflects the number of different ecological units in the ecosystem. By analyzing the total number of patches (NP), we can understand the complexity and diversity of the ecosystem; the average patch area... It reflects the size of each patch in the ecosystem. A larger average patch area usually means that each patch can support more biological populations and ecological processes. Boundary density ED reflects the ratio of the boundary length between different patches in the ecosystem to the total area. A higher boundary density usually means that there are more boundaries between patches in the ecosystem, which may increase the edge effect.

[0105] Figure 3 for Figure 1 A flowchart illustrating the detailed steps of step S1 in an ecological management assessment method 1000 based on ecological patterns. (See attached diagram.) Figure 3 As shown, step S1 includes steps S11-S16.

[0106] In some specific embodiments, in step S1, the dynamic weight allocation model includes a reinforcement learning model, which includes a state space and an action space. The state space enables the agent to perceive the current environmental conditions of the ecosystem and provides a basis for the agent's decision-making. The action space limits the agent's actions, allowing the agent to know which actions are available.

[0107] Optionally, the state space includes ecological connectivity ratio, ecosystem composition ratio, ecosystem type area change rate, ecosystem spatial pattern characteristics, ecosystem comprehensive dynamic degree, ecosystem classification dynamic degree, and ecological planning T1, ecological protection T2, and ecological restoration T3. Optionally, the action space includes a first weight matrix W1, and the value range and adjustment step size of each data in the first weight matrix W1. For example, the value range of the weight data can be set in the interval [0,1], and the adjustment step size for iterative updates of each weight data is 0.01. By reasonably setting the value range and adjustment step size of the weight data, the flexibility and accuracy of weight allocation can be ensured.

[0108] Optionally, in step S1, the state space also includes other environmental variables related to ecological management, such as regional climate conditions and the intensity of human activities, to more comprehensively reflect the actual state of the ecosystem and external influencing factors.

[0109] In some specific embodiments, the action space represents the weight allocation strategy for the six dimensions of data in the ecological pattern evaluation matrix. Optionally, the action space includes a first weight matrix W1, and the value range and adjustment step size of each data in the first weight matrix W1. For example, the weight value range can be set in the interval [0,1], and the weight adjustment step size for each dimension is 0.01, which can ensure the flexibility and accuracy of weight allocation.

[0110] In some specific embodiments, in step S11, the processor designs different reward functions for different ecological management objectives. .

[0111] In some specific embodiments, the ecological management objective is ecological planning T1. The reward function of ecological planning T1 should focus on the rational layout and long-term sustainable development of the ecosystem. The formula of the reward function includes: .in, , , These are the changes in the proportion of ecosystem components. Changes in the rate of change of ecosystem type area Changes in the spatial pattern characteristics of the ecosystem The weighting coefficients.

[0112] In some specific embodiments, the ecological management objective is ecological protection T2, and the reward function for ecological protection T2 should focus on maintaining the stability and connectivity of the ecosystem. The formula for the reward function includes: .in, , , These are the changes in the ecological connectivity ratio. Changes in the proportion of ecosystem components Changes in the rate of change of ecosystem type area The weighting coefficients.

[0113] In some specific embodiments, the ecological management objective is ecological restoration T3, and the reward function for ecological restoration T3 should focus on the recovery speed and quality of the ecosystem. The formula for the reward function includes: .in, , , These are the changes in the area change rate of each ecosystem type. Changes in the overall dynamic degree of the ecosystem Changes in the dynamic degree of ecosystem classification The weighting coefficients.

[0114] In some specific embodiments, in order to enable the reinforcement learning model to adapt to different ecological management objectives, a comprehensive reward function is designed in step S11, and the weights of the reward function are dynamically adjusted according to the current management objective, for example: ,in, , , These are weighting coefficients that are dynamically adjusted based on current ecological management goals, and When the ecological management objective is ecological planning T1, add... The value is reduced. and The value; when the ecological management goal is ecological protection T2, increase The value is reduced. and The value; when the ecological management goal is ecological restoration T3, increase The value is reduced. and The value of .

[0115] In step S12, the processor preprocesses the management operation dataset to obtain training samples. Optionally, the management operation dataset includes historical ecological data under different ecological management objectives and corresponding management operation records, specifically including changes in the six dimensions of the ecological pattern evaluation matrix at different time points, and response results under different management operations (such as the implementation of ecological restoration projects, adjustments to ecological protection policies, etc.). In step S13, the processor performs preprocessing to ensure the quality and consistency of the data, making it suitable as effective training samples input into the reinforcement learning model.

[0116] In step S13, the agent of the reinforcement learning model selects a first weight matrix W1 from the action space as the action to be executed based on the current state. In some specific embodiments, the reinforcement learning model includes a decision-making agent that learns by interacting with the environment. In some specific embodiments, the agent in the reinforcement learning model is placed in a simulated ecological management environment, and the agent selects a weight allocation strategy (i.e., the first weight matrix W1) from the action space as the action to be executed based on the current state (including six-dimensional data, ecological management goals, etc.).

[0117] In step S14, the impact of the selected first weight matrix W1 on the ecological management objective is evaluated according to the reward function, and the reward signal and new state information are fed back to the agent. In some specific embodiments, the reward signal is the change of the six dimensions of the ecological pattern evaluation matrix after weight adjustment, and the state information is the corresponding response of the ecosystem.

[0118] In step S15, the agent updates and learns its weight allocation plan using a reinforcement learning algorithm based on the reward signal and new state information to determine the final reinforcement learning model. In some specific embodiments, the reinforcement learning algorithm includes any one of Q-learning algorithm, deep deterministic policy gradient algorithm, SARSA algorithm, and deep Q network.

[0119] In some specific embodiments, in step S15, through continuous interaction with the environment and trial and error, the agent can gradually learn which weight allocation strategies can obtain higher cumulative rewards under different ecological management goals and ecosystem states, thereby optimizing its weight allocation decisions and ultimately obtaining a reinforcement learning model.

[0120] According to the above implementation scheme, the ecological management assessment method based on ecological patterns proposed in this invention can achieve cross-target transfer learning and dynamic adaptation to environmental changes. Because the reinforcement learning model accumulates experience and knowledge of weight allocation under different ecological management objectives during training, when switching from one ecological management objective to another, the model can utilize some of the previously learned general knowledge and patterns to adapt to the new objective scenario more quickly and find a weight allocation strategy suitable for the new objective in fewer training iterations. The reinforcement learning weight adjustment model has good dynamic adaptability and can adjust the weight allocation strategy in real time according to changes in the state of the ecosystem and interference from external environmental factors.

[0121] Figure 4 for Figure 1 A flowchart illustrating the detailed steps of step S3 in an ecological management assessment method 1000 based on ecological patterns. (See attached diagram.) Figure 4 As shown, step S3 includes steps S31-S33.

[0122] In step S31, the connectivity index COH, the first clustering index AI, the total number of patches NP, and the average patch area are calculated based on the ecological pattern dataset. Boundary density ED, second aggregation index C.

[0123] Optionally, the connectivity index (COH) is calculated using the following formula: ;

[0124] in, Indicates the perimeter of the plaque. Let A represent the area of ​​the patch, and let A represent the total number of patches within the target area.

[0125] Optionally, the first clustering index AI is calculated according to the following formula: ;

[0126] in, This represents the number of adjacent elements of patch type i, and max represents the maximum number of adjacent elements that patch type i can have. This indicates the proportion of the landscape containing type i patches within the target area.

[0127] In some specific embodiments, the processor measures the average patch area. The calculation follows the formula: ;

[0128] Where Aij represents the area of ​​the j-th patch in the i-th ecosystem, and Ni represents the total number of patches in the i-th ecosystem.

[0129] In some specific embodiments, the processor calculates the boundary density ED according to the following formula: ;

[0130] Where Lij represents the boundary length between the i-th type of ecosystem patch and the adjacent j-th type of ecosystem patch, and Ai represents the total area of ​​the i-th type of ecosystem.

[0131] In some specific embodiments, the processor calculates the second aggregation index C according to the following formula:

[0132] Where Pij represents the probability that patch types i and j are adjacent, n represents the total number of patches in each ecosystem, and Cmax represents the maximum value of the Pij parameter.

[0133] In step S32, the ecological connectivity ratio is calculated based on the connectivity index COH and the first aggregation index AI, following the formula: .

[0134] In some specific embodiments, the processor sets the second weight matrix W2 by a preset method. In some specific embodiments, the processor obtains the second weight matrix W2 based on a dynamic weight allocation model. .

[0135] In step S33, based on the total number of patches NP and the average patch area The spatial pattern characteristics of the ecosystem are calculated using the boundary density ED and the second aggregation index C, following the formula: .

[0136] In some specific embodiments, in step S13, the processor sets the third weight matrix in a preset manner. In some specific embodiments, the processor obtains the third weight matrix based on a dynamic weight allocation model. .

[0137] According to such Figure 4 As shown in the embodiments, the technical solution proposed in this invention can obtain the weight values ​​of the weighted summation of refined features during the calculation of ecological connectivity ratios and ecosystem spatial pattern characteristics through a preset or dynamic weight allocation model. The embodiment shown in Figure 3 enables more flexible calculation of ecological connectivity ratios and ecosystem spatial pattern characteristics, and can adapt to different ecological management objectives or other different ecological management requirements.

[0138] Figure 5 for Figure 1 A flowchart illustrating the detailed steps of step S5 in an ecological management assessment method 1000 based on ecological patterns. (See attached diagram.) Figure 5 As shown, step S5 includes steps S51-S54.

[0139] In step S51, if the ecological pattern assessment score SC is less than or equal to the first score, the completion level is poor. In step S52, if the ecological pattern assessment score SC is greater than the first score and less than or equal to the second score, the completion level is average. In step S53, if the ecological pattern assessment score SC is greater than the second score and less than or equal to the third score, the completion level is good. In step S54, if the ecological pattern assessment score SC is greater than the third score, the completion level is excellent.

[0140] In steps S51-S54, the completion level reflects the achievement of ecological management objectives and can serve as a basis for subsequent improvements to ecological management operations. Optionally, the first, second, and third score values ​​are 0.5, 0.6, and 0.8, respectively. Optionally, the values ​​of the first, second, and third score values ​​can be flexibly set according to ecological management objectives or other assessment-related requirements.

[0141] Figure 6 This is a schematic diagram of the structure of an ecological management assessment device 3000 based on ecological patterns according to the present invention. Figure 6 As shown, the device 3000 includes a weight model training module 31, a weight matrix generation module 32, an evaluation matrix generation module 33, a score calculation module 34, and a grade evaluation module 35.

[0142] In some specific embodiments, the weight model training module 31 is trained based on the ecological management goals and management operation dataset to obtain a dynamic weight allocation model. In some specific embodiments, the dynamic weight allocation model is a reinforcement learning model, which includes a state space and an action space.

[0143] In some specific embodiments, the weighted model training module 31 defines the state space of the reinforcement learning model. Optionally, the state space includes ecological connectivity ratios, ecosystem composition ratios, ecosystem type area change rates, ecosystem spatial pattern characteristics, ecosystem comprehensive dynamics, ecosystem classification dynamics, and ecological planning T1, ecological protection T2, and ecological restoration T3. Optionally, the state space also includes other environmental variables related to ecological management, such as regional climate conditions and the intensity of human activities, to more comprehensively reflect the actual state of the ecosystem and external influencing factors.

[0144] In some specific embodiments, the weight model training module 31 defines the action space of the reinforcement learning model, which represents the weight allocation strategy for the six dimensions of data in the ecological pattern evaluation matrix. Optionally, the action space includes a first weight matrix W1, and the value range and adjustment step size of each data in the first weight matrix W1. For example, the value range of the weight data can be set in the interval [0,1], and the adjustment step size for iteratively updating each weight data is 0.01. By reasonably setting the value range and adjustment step size of the weight data, the flexibility and accuracy of weight allocation are ensured.

[0145] In some specific embodiments, the weighted model training module 31 sets different reward functions for different ecological management objectives. Specifically, the reward function The corresponding ecological management objectives are ecological planning (T1), ecological protection (T2), and ecological restoration (T3).

[0146] In some specific embodiments, the weighted model training module 31 preprocesses the data in the management operation dataset to obtain training samples. The preprocessing operations specifically include data cleaning and data normalization. Optionally, the management operation dataset includes historical ecological data under different ecological management objectives and corresponding management operation records, specifically including changes in the six dimensions of the ecological pattern evaluation matrix at different time points, and response results under different management operations (such as the implementation of ecological restoration projects and adjustments to ecological protection policies).

[0147] In some specific embodiments, the weight model training module 31 preprocesses the management operation dataset to obtain training samples; the agent of the reinforcement learning model selects a first weight matrix W1 from the action space as an action to be executed according to the current state; the weight model training module 31 evaluates the impact of the selected first weight matrix W1 on the ecological management goal according to the reward function, and feeds back the reward signal and new state information to the agent; the agent updates and learns its weight allocation plan using the reinforcement learning algorithm according to the reward signal and new state information, so as to determine the final reinforcement learning model.

[0148] In some specific embodiments, the dynamic weight allocation model is a linear regression model. In some specific embodiments, it is assumed that there is a linear relationship between the achievement of ecological management goals and the six dimensions of data in the ecological pattern evaluation matrix; that is, the goal achievement score can be represented as a linear combination of the six dimensions. Using a linear regression algorithm, historical data from the management operation dataset is used, with the six dimensions as independent variables and the actual achievement of ecological management goals as the dependent variable, to train the model and obtain the regression coefficients corresponding to each dimension. These regression coefficients can then serve as the mapping matrix between ecological management goals and weight allocation.

[0149] In some specific embodiments, the dynamic weight allocation model is an artificial neural network model. In some specific embodiments, the complex mapping relationship between the six dimensions of data and the achievement of ecological management goals is learned through training. During training, the connection weights of the neural network are continuously adjusted, eventually converging to weight values ​​that can better predict the achievement of goals. These weight values ​​can reflect the importance of the six dimensions of data in the ecological pattern evaluation matrix.

[0150] In some specific embodiments, the weight matrix generation module 32 determines the first weight matrix according to the dynamic weight allocation model. In some specific embodiments, the weight matrix generation module 32 can obtain the specific values ​​of the first weight matrix W1 corresponding to the ecological management objective based on the already trained reinforcement learning model. For example, in some specific embodiments, the ecological management objective is ecological planning T1, and the first weight matrix... The ecological management objective is ecological protection T2, with the first weight matrix. The ecological management objective is ecological restoration T3, with the first weight matrix. .

[0151] In some specific embodiments, the evaluation matrix generation module 33 determines the ecological pattern evaluation matrix based on the ecological pattern dataset of the target area. .

[0152] in, Indicates the ratio of ecological connectivity. Indicates the proportion of ecosystem components. Indicates the rate of change in the area of ​​ecosystem types. Indicates the spatial pattern characteristics of the ecosystem. Indicates the overall dynamic degree of the ecosystem. This indicates the dynamic degree of ecosystem classification.

[0153] In some specific embodiments, the evaluation matrix generation module 33 calculates the connectivity index COH and the first clustering index AI based on the ecological pattern dataset, and further calculates the ecological connectivity ratio based on the connectivity index COH and the first clustering index AI. .

[0154] In some specific embodiments, the evaluation matrix generation module 33 sets the second weight matrix W2= using a preset method. .

[0155] In some specific embodiments, the evaluation matrix generation module 33 obtains the second weight matrix W2 based on the dynamic weight allocation model. .

[0156] In some specific embodiments, the evaluation matrix generation module 33 calculates the ecological connectivity ratio from the connectivity index COH and the first aggregation index AI, following the formula: .

[0157] In some specific embodiments, the connectivity index (COH) primarily reflects the connectivity between different ecological units within an ecosystem, such as the degree of physical connection between habitat patches, while the first aggregation index (AI) focuses on reflecting the spatial aggregation of ecological elements. By combining the two to calculate the ecological connectivity ratio, the connectivity status of an ecosystem can be assessed more comprehensively and accurately. For example, an ecosystem may have a high connectivity, indicating numerous connection channels between habitat patches. However, if these patches have low aggregation and are scattered, the overall connectivity of the ecosystem may not be ideal. The ecological connectivity ratio can comprehensively reflect this complex situation.

[0158] In some specific embodiments, the evaluation matrix generation module 33 calculates the ecosystem composition ratio based on the ecological pattern dataset, following the formula: ;

[0159] in, TS represents the total area of ​​the target area being assessed, where i represents the area of ​​ecosystem type i.

[0160] In some specific embodiments, the evaluation matrix generation module 33 calculates the ecosystem area change rate based on the ecological pattern dataset, following the formula: ;in, This represents the initial area of ​​ecosystem type i. This represents the area of ​​ecosystem type i at the end of the period.

[0161] In some specific embodiments, the evaluation matrix generation module 33 calculates the total number of patches NP and the average patch area based on the ecological pattern dataset. The boundary density ED and the second aggregation index C are used to further calculate the spatial pattern characteristics of the ecosystem.

[0162] In some specific embodiments, the weight matrix generation module 32 sets the third weight matrix in a preset manner. In some specific embodiments, the weight matrix generation module 32 obtains the third weight matrix according to the dynamic weight allocation model. .

[0163] In some specific embodiments, the evaluation matrix generation module 33 evaluates the total number of patches NP and the average patch area. The spatial pattern characteristics of the ecosystem are calculated using the boundary density ED and the second aggregation index C, following the formula: .

[0164] In some specific embodiments, the evaluation matrix generation module 33 calculates the comprehensive dynamic degree of the ecosystem based on the ecological pattern dataset, following the formula: Where ECOi represents the initial area of ​​ecosystem type i. This represents the absolute value of the area where ecosystem type i is converted to a non-i type ecosystem.

[0165] In some specific embodiments, the evaluation matrix generation module 33 calculates the ecosystem classification dynamics based on the ecological pattern dataset, following the formula: Where i represents the initial ecosystem type and j represents the final ecosystem type. Let Aij represent the area of ​​each ecosystem type, Aij represent the proportion of the i-th ecosystem type at the beginning of the period that transformed into the j-th ecosystem type at the end of the period, and Biij represent the proportion of the j-th ecosystem type at the end of the period that transformed from the i-th ecosystem type at the beginning of the period.

[0166] In some specific embodiments, the score calculation module 34 determines the ecological pattern assessment score SC = W1 × D1 based on the first weight matrix W1 and the ecological pattern evaluation matrix D1. In some specific embodiments, the ecological pattern assessment score SC = .

[0167] In some specific embodiments, the rating module 35 determines the completion level of the ecological management objectives based on the ecological pattern assessment score SC. In some specific embodiments, if the ecological pattern assessment score SC is less than or equal to a first score, the completion level is poor; if the ecological pattern assessment score SC is greater than the first score and less than or equal to a second score, the completion level is average; if the ecological pattern assessment score SC is greater than the second score and less than or equal to a third score, the completion level is good; and if the ecological pattern assessment score SC is greater than the third score, the completion level is excellent.

[0168] Optionally, the first, second, and third rating values ​​are 0.5, 0.6, and 0.8, respectively.

[0169] Figure 7 This is a schematic diagram of the structure of an integrated ecological monitoring system 4000 according to the present invention. Figure 7 As shown, the system 4000 includes a data acquisition unit 41, a data analysis unit 42, and a data visualization unit 43.

[0170] In some specific embodiments, the data acquisition unit 41 includes a satellite remote sensing monitoring module, a drone aerial photography monitoring module, and an Internet of Things (IoT) acquisition module, used to collect ecological pattern datasets and management operation datasets. The satellite remote sensing monitoring module, the drone aerial photography monitoring module, and the IoT acquisition module work together to monitor the ecological conditions of the target area, such as soil change trends, water flow changes, and forest changes, and generate ecological pattern datasets and management operation datasets from the collected data.

[0171] Among them, the satellite remote sensing monitoring module conducts meteorological monitoring through satellite imaging and remote sensing data, the drone aerial photography monitoring monitors organisms, soil, patches and various biodiversity components, and the Internet of Things (IoT) acquisition module includes a variety of network-connected sensing devices. The IoT acquisition module can directly monitor data on substances such as atmosphere, soil and water quality in the target area.

[0172] In some specific embodiments, the data analysis unit 42 includes, for example: Figure 6 The ecological management assessment device based on ecological patterns shown is used to perform tasks such as... Figure 1-5 The method described by either of them.

[0173] In some specific embodiments, the data visualization unit 43 includes a visualization data display platform for displaying the analysis results of the data analysis unit. In some specific embodiments, the visualization data display platform includes an interactive page and a processor, capable of intuitively displaying the management status of the ecosystem based on the evaluation results of the data unit 42, and generating corresponding tables, charts, reports, etc.

[0174] The embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, any changes or modifications made by those skilled in the art based on the ideas of the present invention, its specific implementation methods, and its application scope, are all within the scope of protection of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An ecological management assessment method based on ecological patterns, characterized in that, include: S1. A dynamic weight allocation model is obtained by training on an ecological management goal and management operation dataset; S2. Determine the first weight matrix according to the dynamic weight allocation model. ; S3. Determine the ecological pattern evaluation matrix based on the ecological pattern dataset of the target area. ; in, Indicates the ratio of ecological connectivity. Indicates the proportion of ecosystem components. Indicates the rate of change in the area of ​​ecosystem types. Indicates the spatial pattern characteristics of the ecosystem. Indicates the overall dynamic degree of the ecosystem. Indicates the dynamic degree of ecosystem classification; S4. Determine the ecological pattern assessment score SC = W1 × D1 based on the first weight matrix W1 and the ecological pattern evaluation matrix D1; S5. Determine the completion level of the ecological management objectives based on the ecological pattern assessment score SC; The ecological management objective is any one of ecological planning T1, ecological protection T2, and ecological restoration T3. The dynamic weight allocation model includes a reinforcement learning model, which includes a state space and an action space. The state space records ecosystem parameters, including the ecological connectivity ratio, ecosystem composition ratio, ecosystem type area change rate, ecosystem spatial pattern characteristics, ecosystem comprehensive dynamic degree, ecosystem classification dynamic degree, as well as ecological planning T1, ecological protection T2, and ecological restoration T3; it is used to enable the agent to perceive the current environmental status of the ecosystem and to provide a basis for the agent's decision-making. The action space includes the first weight matrix W1, and the value range and adjustment step size of each data in the first weight matrix W1, which are used to limit the action mode of the agent and enable the agent to know what actions are available. The S1. dynamic weight allocation model, trained based on the management operation dataset, includes: Different reward functions are set for different ecological management objectives. Specifically, this includes: the ecological management objective is the ecological plan T1, and the reward function... The ecological management objective is the ecological protection T2, and the reward function is... The ecological management objective is the ecological restoration T3, and the reward function is... ;in, , , , , , These are the changes in the ecological connectivity ratio. The change in the proportion of the ecosystem components The change value of the area change rate of the ecosystem type The change values ​​of the spatial pattern characteristics of the ecosystem The change value of the overall dynamic degree of the ecosystem. The change value of the dynamic degree of ecosystem classification. Weighting coefficients; Further design of the comprehensive reward function ;in, , , These are weighting coefficients that are dynamically adjusted based on the stated ecological management objectives, and When the ecological management objective is ecological planning T1, add... The value is reduced. and The value; when the ecological management objective is ecological protection T2, increase The value is reduced. and The value; when the ecological management objective is ecological restoration T3, increase The value is reduced. and The value; The management operation dataset is preprocessed to obtain training samples; The agent in the reinforcement learning model selects a first weight matrix W1 from the action space as an action to execute based on the current state; The impact of the selected first weight matrix W1 on the ecological management objective is evaluated according to the reward function, and the reward signal and new state information are fed back to the agent. The agent updates and learns its weight allocation plan using a reinforcement learning algorithm based on reward signals and new state information to determine the final reinforcement learning model. The state space also includes the connectivity index COH, the first aggregation index AI, the total number of patches NP, and the average patch area. The action space also includes a boundary density ED, a second clustering index C, and a second weight matrix. and the third weight matrix The ecological management assessment method based on ecological patterns also includes: The second weight matrix is ​​determined based on the dynamic weight allocation model. and the third weight matrix .

2. The ecological management assessment method based on ecological pattern as described in claim 1, characterized in that, The agent in the reinforcement learning model is the core component for decision-making and learning. It is used to select appropriate actions from its action space to perform by perceiving the current state of the environment, and then adjust its behavioral strategy according to the reward signals fed back by the environment.

3. The ecological management assessment method based on ecological pattern as described in claim 1, characterized in that, S3. Determining the ecological pattern evaluation matrix D1 based on the ecological pattern dataset of the target area includes: Based on the aforementioned ecological pattern dataset, the connectivity index (COH), the first clustering index (AI), the total number of patches (NP), and the average patch area were calculated. Boundary density ED, second aggregation index C; The ecological connectivity ratio is calculated based on the connectivity index (COH) and the first aggregation index (AI), following the formula: ; Based on the total number of patches NP and the average patch area The spatial pattern characteristics of the ecosystem are calculated using the boundary density ED and the second aggregation index C, following the formula: 。 4. The ecological management assessment method based on ecological pattern as described in any one of claims 1-3, characterized in that, S5. Determining the completion level of the ecological management objectives based on the ecological pattern assessment score SC includes: If the ecological pattern assessment score SC is less than or equal to the first score, the completion level is poor. If the ecological pattern assessment score SC is greater than the first score value and less than or equal to the second score value, the completion level is considered average. If the ecological pattern assessment score SC is greater than the second score and less than or equal to the third score, the completion level is good. If the ecological pattern assessment score SC is greater than the third score, the completion status is rated as excellent.

5. An ecological management assessment device based on ecological patterns, characterized in that, include: The evaluation matrix generation module is used to determine the ecological pattern evaluation matrix based on the ecological pattern dataset of the target area. ; in, Indicates ecological connectivity. Indicates the proportion of ecosystem components. Indicates the rate of change in the area of ​​ecosystem types. Indicates the spatial pattern characteristics of the ecosystem. Indicates the overall dynamic degree of the ecosystem. Indicates the dynamic degree of ecosystem classification; The weight model training module is used to train a dynamic weight allocation model based on the ecological management goals and management operation dataset. The weight matrix generation module is used to determine the first weight matrix based on the dynamic weight allocation model. ; The scoring module is used to determine the ecological pattern assessment score SC = W1 × D1 based on the first weight matrix W1 and the ecological pattern evaluation matrix D1. The rating module is used to determine the completion level of the ecological management objectives based on the ecological pattern assessment score (SC). The ecological management objective is any one of ecological planning T1, ecological protection T2, and ecological restoration T3. The dynamic weight allocation model includes a reinforcement learning model, which includes a state space and an action space. The state space records ecosystem parameters, including the ecological connectivity ratio, ecosystem composition ratio, ecosystem type area change rate, ecosystem spatial pattern characteristics, ecosystem comprehensive dynamic degree, ecosystem classification dynamic degree, as well as ecological planning T1, ecological protection T2, and ecological restoration T3; it is used to enable the agent to perceive the current environmental status of the ecosystem and to provide a basis for the agent's decision-making. The action space includes the first weight matrix W1, and the value range and adjustment step size of each data in the first weight matrix W1, which are used to limit the action mode of the agent and enable the agent to know what actions are available. The weight model training module, based on the ecological management goals and management operation dataset, specifically includes the following process for obtaining the dynamic weight allocation model: Different reward functions are set for different ecological management objectives. Specifically, this includes: the ecological management objective is the ecological plan T1, and the reward function... The ecological management objective is the ecological protection T2, and the reward function is... The ecological management objective is the ecological restoration T3, and the reward function is... ;in, , , , , , These are the changes in the ecological connectivity ratio. The change in the proportion of the ecosystem components The change value of the area change rate of the ecosystem type The change values ​​of the spatial pattern characteristics of the ecosystem The change value of the overall dynamic degree of the ecosystem. The change value of the dynamic degree of ecosystem classification. Weighting coefficients; Further design of the comprehensive reward function ;in, , , These are weighting coefficients that are dynamically adjusted based on the stated ecological management objectives, and When the ecological management objective is ecological planning T1, add... The value is reduced. and The value; when the ecological management objective is ecological protection T2, increase The value is reduced. and The value; when the ecological management objective is ecological restoration T3, increase The value is reduced. and The value; The management operation dataset is preprocessed to obtain training samples; The agent in the reinforcement learning model selects a first weight matrix W1 from the action space as an action to execute based on the current state; The impact of the selected first weight matrix W1 on the ecological management objective is evaluated according to the reward function, and the reward signal and new state information are fed back to the agent. The agent updates and learns its weight allocation plan using a reinforcement learning algorithm based on reward signals and new state information to determine the final reinforcement learning model. The state space also includes the connectivity index COH, the first aggregation index AI, the total number of patches NP, and the average patch area. The action space also includes a boundary density ED, a second clustering index C, and a second weight matrix. and the third weight matrix The ecological management assessment method based on ecological patterns also includes: The second weight matrix is ​​determined based on the dynamic weight allocation model. and the third weight matrix .

6. An integrated ecological monitoring system, characterized in that, include: The data acquisition unit includes a satellite remote sensing monitoring module, a drone aerial photography monitoring module, and an Internet of Things acquisition module, which are used to collect ecological pattern datasets and management operation datasets. The data analysis unit includes the ecological management assessment device based on ecological pattern as described in claim 5, for executing the ecological management assessment method based on ecological pattern as described in any one of claims 1-4; The data visualization unit includes a data visualization platform for displaying the analysis results of the data analysis unit.

Images