Intelligent agricultural management method and system based on image processing

Abstract

The invention provides an intelligent agricultural management method and system based on image processing, and the method comprises the steps: carrying out the rice-fish symbiotic region network deployment of a rice-fish symbiotic region, carrying out the multi-source data collection of rice, fish and environment based on the rice-fish symbiotic region network, obtaining the rice-fish symbiotic ecological data, carrying out the data preprocessing of the rice-fish symbiotic ecological data, and carrying out the data processing of the rice-fish symbiotic ecological data. Constructing a rice-fish data heterogeneous graph based on the rice-fish symbiotic ecological data after data preprocessing, constructing a rice-fish attention network model, importing the rice-fish data heterogeneous graph into the rice-fish attention network model, performing cross-domain feature fusion on the rice-fish symbiotic ecological data based on the rice-fish attention network model, and obtaining rice-fish associated data. And then performing rice-fish causal reasoning based on the rice-fish association data, generating a rice-fish association knowledge graph, establishing a rice-fish decision model, importing the rice-fish association knowledge graph into the rice-fish decision model, and generating a rice-fish regulation and control decision. Through the method, an accurate and efficient intelligent dynamic regulation and control decision is provided for a user.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural management, and more particularly, to an intelligent agricultural management method and system based on image processing. Background Art

[0002] As a typical ecological agriculture model, the rice-fish symbiotic system realizes resource recycling and efficiency improvement through the synergy of rice planting and aquaculture. However, the traditional rice-fish symbiotic system has limitations when applied to this scenario.

[0003] On the one hand, the data monitoring system of the traditional rice-fish symbiotic system mostly focuses on single elements and lacks multi-dimensional collaborative perception of rice, fish, and the environment. For example, although the fish farming system regulates the feeding machine through a dissolved oxygen sensor, it is not linked to the dynamic nitrogen demand during the growth period of rice, resulting in waste of nitrogen cycle resources. Moreover, the data analysis method relies on shallow statistical correlations, simply establishing a linear causal hypothesis between water temperature and fish feeding amount, while ignoring the chain effect between organisms, making the model often unable to describe the non-linear biological interaction process, thus affecting the generation of subsequent decisions.

[0004] On the other hand, at the decision-making level, the traditional rice-fish symbiotic system relies on a pre-set rule library and adopts a single-threshold trigger mechanism, neither considering the synergy effect of environmental parameters nor being compatible with the lag response of biological behaviors. For example, under continuous rainy conditions, excessive oxygenation may exacerbate heat dissipation due to water body disturbance, resulting in a sudden drop in water temperature, which instead inhibits fish activity. This contradictory scenario exposes the fundamental defect of the traditional system's lack of multi-dimensional parameter dynamic weight calculation ability, being unable to quantify the gradient relationship between oxygenation intensity and water temperature change, nor being able to predict the cross-scale feedback impact of regulation measures on the symbiotic system, thus leading to rigid decision-making.

[0005] Therefore, the existing traditional rice-fish symbiotic system technology is unable to cope with the complex and changeable rice-fish symbiotic system, and cannot provide precise and comprehensive decisions to meet the refined requirements of the rice-fish symbiotic system. Summary of the Invention

[0006] In view of the problems mentioned above, in combination with the first aspect of the present invention, embodiments of the present invention provide an intelligent agricultural management method based on image processing, the method comprising: Deploy a rice-fish symbiotic area network for the rice-fish symbiotic area, where the rice-fish symbiotic area refers to the area where rice is planted while fish are cultured; Collect multi-source data of rice, fish, and the environment based on the rice-fish symbiotic area network to obtain rice-fish symbiotic ecological data, where the rice-fish symbiotic ecological data includes rice data, fish data, and environmental data, and perform data preprocessing on the rice-fish symbiotic ecological data; Construct a heterogeneous graph of rice-fish data based on the rice-fish symbiotic ecological data after data preprocessing; Construct a rice-fish attention network model, import the heterogeneous graph of rice-fish data into the rice-fish attention network model, perform cross-domain feature fusion on the rice-fish symbiotic ecological data based on the rice-fish attention network model, and obtain rice-fish associated data; Perform rice-fish causal reasoning based on the rice-fish associated data to generate a rice-fish associated knowledge graph; Establish a rice-fish decision-making model, import the rice-fish associated knowledge graph into the rice-fish decision-making model, and generate rice-fish regulation decisions. Users can perform dynamic regulation on the rice-fish symbiotic area based on the rice-fish regulation decisions.

[0007] As a further solution of the present invention, the importing the heterogeneous graph of rice-fish data into the rice-fish attention network model, performing cross-domain feature fusion on the rice-fish symbiotic ecological data based on the rice-fish attention network model, and obtaining rice-fish associated data includes: Import the preprocessed rice-fish symbiotic ecological data and the heterogeneous graph of rice-fish data into the rice-fish attention network model, obtain the feature vectors of the rice-fish symbiotic ecological data based on the rice-fish attention network model, perform spatial projection on the feature vectors of the rice-fish symbiotic ecological data, and map the feature vectors of the rice-fish symbiotic ecological data to the same space; Obtain the ecological attention weights of the nodes included in the heterogeneous graph of rice-fish data, perform rice-fish association discovery on the feature vectors of the rice-fish symbiotic ecological data based on the ecological attention weights, and obtain rice-fish associated data.

[0008] As a further solution of the present invention, the constructing the rice-fish attention network model includes: When constructing the rice-fish attention network model, import the dynamic meta-learning framework into the rice-fish attention network model; Divide the rice-fish symbiotic area into block partitions, divide the rice-fish symbiotic area into equal-area rectangular sub-regions, and obtain the historical rice-fish symbiotic ecological data of each rectangular sub-region; Use the historical rice-fish symbiotic ecological data of the same rectangular sub-region in different periods as training positive samples, randomly combine the historical rice-fish symbiotic ecological data of different rectangular sub-regions in the same period, and use the randomly combined historical rice-fish symbiotic ecological data as training negative samples; Train the rice-fish attention network model based on the training positive samples and the training negative samples.

[0009] As a further solution of the present invention, the performing rice-fish causal reasoning based on the rice-fish associated data and the heterogeneous graph of rice-fish data to generate a rice-fish associated knowledge graph includes: Optimize the PC calculation for rice-fish association data, and during the process of optimizing the PC calculation, add time lag constraints and ecological prior constraints. The time lag constraints are used to constrain the reasons for causal inference, and the ecological prior constraints are used to constrain the inference logic of causal inference. Based on the optimized PC calculation, obtain the linear causal relationship between rice and fish, perform non-linear expansion on the non-Gaussian distributed data in the rice-fish association data, identify and obtain the non-linear causal relationship in the rice-fish association data, and generate an initial rice-fish causal network based on the non-linear causal relationship and the linear causal relationship; Calculate the rice-fish treatment effect data of rice and fish, and construct a rice-fish association knowledge graph based on the initial rice-fish causal network and the rice-fish treatment effect data of rice and fish.

[0010] As a further solution of the present invention, the establishment of the rice-fish decision model, importing the rice-fish association knowledge graph into the rice-fish decision model, and generating a rice-fish regulation decision, includes: Construct a rice-fish decision space based on the rice-fish association knowledge graph and combined with the rice-fish symbiotic ecological data, perform rice-fish ecological prediction according to the initial rice-fish causal network included in the rice-fish association knowledge graph, and generate a rice-fish regulation decision based on the result of the rice-fish ecological prediction. The rice-fish regulation decision includes an explicit regulation decision and an implicit regulation decision.

[0011] As a further solution of the present invention, the construction of the rice-fish decision space based on the rice-fish association knowledge graph and combined with the rice-fish symbiotic ecological data includes: Obtain the rice knowledge nodes, fish knowledge nodes, and environmental knowledge nodes in the rice-fish association knowledge graph, obtain the knowledge node data based on the rice knowledge nodes, fish knowledge nodes, and environmental knowledge nodes, convert the knowledge node data into rice-fish decision space vectors, and convert the initial rice-fish causal network included in the rice-fish association knowledge graph into a rice-fish association weight matrix; Construct a rice-fish decision space based on the rice-fish decision space vectors and the rice-fish association weight matrix.

[0012] As a further solution of the present invention, the construction of the rice-fish data heterogeneous graph based on the preprocessed rice-fish symbiotic ecological data includes: Define the graph nodes for the preprocessed rice-fish symbiotic ecological data, obtain the rice heterogeneous nodes, fish heterogeneous nodes, and environmental heterogeneous nodes, add edge relationship rules to the rice heterogeneous nodes, fish heterogeneous nodes, and environmental heterogeneous nodes, and link the rice heterogeneous nodes, fish heterogeneous nodes, and environmental heterogeneous nodes based on the edge relationship rules to generate a rice-fish data heterogeneous graph.

[0013] As a further solution of the present invention, the multi-source data of rice, fish and the environment are collected based on the rice-fish symbiotic area network to obtain rice-fish symbiotic ecological data, and the rice-fish symbiotic ecological data includes rice data, fish data and environmental data, including: The rice-fish symbiotic ecological data is obtained by the rice data acquisition terminal, fish data acquisition terminal and environmental data acquisition terminal based on the rice-fish symbiotic area network. The rice data includes chlorophyll content and canopy nitrogen accumulation. The fish data includes fish movement trajectories, fish feeding frequencies and fish population densities. The environmental data includes dissolved oxygen concentration in water, nitrogen ammonia concentration in water, pH value of water and water temperature.

[0014] On the other hand, the embodiment of the present invention also provides an intelligent agricultural management system based on image processing, including: A rice-fish symbiotic area network, which is used to collect rice data, fish data and environmental data in the rice-fish symbiotic area to obtain rice-fish symbiotic ecological data; A composition module, which is used to preprocess the rice-fish symbiotic ecological data and construct a heterogeneous graph of rice-fish data; A rice-fish model module, which is used to construct a rice-fish attention network model and obtain rice-fish association data based on the heterogeneous graph of rice-fish data and the rice-fish symbiotic ecological data; A knowledge construction module, which is used to perform rice-fish causal reasoning and generate a rice-fish association knowledge graph; A rice-fish decision-making module, which is used to generate rice-fish regulation decisions, and users can dynamically regulate the rice-fish symbiotic area based on the rice-fish regulation decisions.

[0015] As a further solution of the present invention, the rice-fish symbiotic area network includes: A rice data acquisition terminal, which includes a drone cluster. The drone cluster is used to cruise the rice-fish symbiotic area with a cruise frequency of twice a day, and obtain the chlorophyll content and canopy nitrogen accumulation based on the drone cluster; A fish data acquisition terminal, which includes an underwater sonar array and an infrared module. The fish are monitored based on the underwater sonar array and the infrared module to obtain fish movement trajectories, fish feeding frequencies and fish population densities; An environmental data acquisition terminal, which is used to monitor the environment of the rice-fish symbiotic area to obtain the dissolved oxygen concentration in water, nitrogen ammonia concentration in water, pH value of water and water temperature.

[0016] Based on the above aspects, in the embodiments of the present application, a rice-fish symbiotic ecological data composed of multi-source acquisition of rice, fish, and environmental data in the rice-fish symbiotic area is deployed through the rice-fish symbiotic area network. And through data preprocessing, the abnormal data in the rice-fish symbiotic ecological data is cleaned to reduce the impact of abnormal data on the accuracy of the rice-fish symbiotic ecological data. By combining the rice-fish attention network and causal reasoning, the complex ecological relationships in the rice-fish symbiotic area are analyzed and quantified, and the causal connections between the abstract and complex rice and fish, rice and environment, and fish and environment are streamlined into specific causal quantification sets. Then, the real-time rice-fish symbiotic ecological data is combined with the causal quantification sets to generate precise dynamic regulation decisions that meet the user's needs. In summary, this intelligent agricultural management method improves the technical limitations of fragmented monitoring, shallow analysis, and static decision-making in the traditional rice-fish symbiotic system, and provides more precise regulation decisions for the intelligent management of the rice-fish symbiotic system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 FIG. is a schematic flowchart of the execution process of an intelligent agricultural management method based on image processing provided by an embodiment of the present invention.

[0018] Figure 2 FIG. is a schematic diagram of an intelligent agricultural management system based on image processing provided by an embodiment of the present invention.

[0019] Figure 3 FIG. is a schematic diagram of the rice-fish symbiotic area network in an intelligent agricultural management system based on image processing provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The present invention will be specifically described below with reference to the accompanying drawings of the specification. Figure 1 FIG. is a schematic flowchart of the execution process of an intelligent agricultural management method based on image processing provided by an embodiment of the present invention. The following provides a detailed introduction to this intelligent agricultural management method based on image processing.

[0021] Step S1, deploy a rice-fish symbiotic area network for the rice-fish symbiotic area, where the rice-fish symbiotic area refers to the rice-fish symbiotic area where rice is planted while fish are cultured.

[0022] Step S2, based on the rice-fish symbiotic area network, perform multi-source data collection on rice, fish, and the environment to obtain rice-fish symbiotic ecological data. The rice-fish symbiotic ecological data includes rice data, fish data, and environmental data, and perform data preprocessing on the rice-fish symbiotic ecological data.

[0023] In this embodiment, step S2 includes: Step S21, based on the rice-fish symbiotic area network, perform multi-source data collection on rice, fish, and the environment to obtain rice-fish symbiotic ecological data.

[0024] Specifically, based on the rice data collection terminal in the rice-fish symbiosis area network, rice data in the rice-fish symbiosis area is obtained, and the rice data includes chlorophyll content and canopy nitrogen accumulation. Based on the fish data collection terminal in the rice-fish symbiosis area network, fish data in the rice-fish symbiosis area is obtained, and the fish data includes fish movement trajectory, fish feeding frequency and fish group density. Based on the environmental data collection terminal in the rice-fish symbiosis area network, environmental data in the rice-fish symbiosis area is obtained, and the environmental data includes dissolved oxygen concentration, nitrogen and ammonia concentration, pH value and water temperature in the rice-fish symbiosis area.

[0025] Step S22, preprocessing the rice-fish symbiosis ecological data.

[0026] In this embodiment, step S22 includes: Step S22-1, performing outlier detection and processing on rice-fish symbiosis ecological data.

[0027] Specifically, the sliding window size is set to 1 hour, and the dynamic mean of similar sensors is monitored in real time. If the sensor data deviates from the mean by more than 20% for three consecutive times and lasts for 10 minutes, it is determined whether the sensor is faulty. For the faulty sensor, redundant switching is used to process the abnormal data. For example, if the data uploaded by pH sensor A1 for three consecutive times exceeds 20% of the dynamic mean of similar sensors, and it is confirmed that the pH sensor is faulty after 10 minutes, the pH sensor data within 5 meters of the pH sensor and the historical data of the pH sensor are obtained, and weighted summation is performed to obtain the corrected data, the data uploaded by the abnormal sensor is shielded, and the corrected data is used to perform data smoothing transition on the abnormal data by means of sliding window linear interpolation. At the same time, the maintenance information sheet is pushed to the maintenance personnel for maintenance. After the maintenance personnel confirms that the maintenance is completed, the value of pH sensor A1 is obtained again, and the value of pH sensor A1 is verified for consistency with the corrected data. After confirming the consistency, the data shielding of pH sensor A1 is cancelled.

[0028] Step S22-2, standardizing the rice-fish symbiosis ecological data.

[0029] Specifically, the environmental data in the rice-fish symbiotic ecological data is standardized based on standardized operations. For example, on a certain day, the water temperature is 27 degrees Celsius, the quarterly average temperature for the quarter in which that day falls is 26 degrees Celsius, and the quarterly standard deviation is 3.1. After standardization, the water temperature on that day is 0.32. Obtain the dissolved oxygen concentration on that day, and calculate the dissolved oxygen offset on that day based on the Z-score algorithm. Standardize the rice data in the rice-fish symbiotic ecological data based on normalization operations. For example, the original chlorophyll value of rice in a certain area is 55 SPAD, and the standardized chlorophyll value is 0.58. Standardize the fish data in the rice-fish symbiotic ecological data based on standardized operations. For example, the fish population density in a certain area is 150 fish per mu, and the data is standardized by logarithmic transformation. The standardized fish population density data is 2.18. Obtain the underwater sonar array data in that area, and calculate the underwater sonar array data based on Shannon entropy to obtain the fish movement trajectory entropy, which is used to reflect the disorder degree of fish group movement.

[0030] Furthermore, calculate the standardization parameters separately according to the tillering stage, heading stage, and maturity stage of rice. For example, the quarterly average water temperature in the tillering stage of rice in a certain rice-fish symbiotic area is 24 degrees Celsius, and the quarterly average water temperature in the heading stage is 28 degrees Celsius. Different standardization parameters are used for data standardization in different quarters to avoid deviation of standardized data parameters due to seasonal deviation, and add the data of adjacent areas in the calculation area to the local mean calculation with weights changing with distance, so as to enhance the accuracy of local mean calculation.

[0031] Step S3: Construct a heterogeneous graph of rice-fish data based on the preprocessed rice-fish symbiotic ecological data.

[0032] In this embodiment, step S3 includes: Step S31: Define heterogeneous nodes based on the rice-fish symbiotic ecological data.

[0033] Specifically, the rice-fish symbiotic area is divided into multiple 5m×5m rice grid units according to a grid, and the rice data within each rice grid unit is used as a rice heterogeneous node. Each rice heterogeneous node contains three node attributes: chlorophyll content, canopy nitrogen accumulation, and the spatial coordinates of the grid unit. For example, the node attributes of rice node A01 are [chlorophyll = 0.65, nitrogen accumulation = 0.44, coordinates (15982, 35746)]. Based on the distribution of the underwater sonar array in the fish data acquisition terminal within the rice-fish symbiotic area network, the rice-fish symbiotic area is divided into multiple 5m×5m water area blocks, and each water area block is used as a fish heterogeneous node. Each fish heterogeneous node includes three node attributes: fish population density, fish feeding frequency, fish movement trajectory entropy, and spatial coordinates. For example, the node attributes of fish node B32 are [density = 2.5, feeding = 13 times / 10 minutes, trajectory entropy = 1.8, coordinates (13795, 24658)]. Each sensor group in the environmental data acquisition terminal within the rice-fish symbiotic area network is used as an environmental heterogeneous node. Each environmental heterogeneous node includes four node attributes: dissolved oxygen offset, ammonia nitrogen concentration, water temperature, and pH value. For example, the node attributes of environmental node C13 are [dissolved oxygen concentration offset = -1.2, ammonia nitrogen concentration = 0.8, water temperature = 0.6, pH = 0.4].

[0034] Step S32, establish the rules for the edge relationships of the heterogeneous graph.

[0035] Specifically, when the distance between the center of the rice node and the sensor group within the environmental node does not exceed 10 meters, a two-way edge is established between the rice node and the environmental node. When the distance between the sonar array within the fish node and the sensor group within the environmental node does not exceed 15 meters, a two-way edge is established. Set the symbiotic relationships among the rice data, fish data, and environmental data. For example, there is a positive correlation between the chlorophyll content of rice and the dissolved oxygen offset in the water body, and there is a dynamic response relationship between the feeding frequency of fish and the water temperature. Set the competition relationships among the rice data, fish data, and environmental data. For example, when the ammonia nitrogen concentration in the water body is greater than 0.5 mg / L, a resource competition relationship is established between rice and fish.

[0036] Furthermore, if it is detected that the significant decrease in the dissolved oxygen concentration in the water body leads to a significant increase in the fish movement trajectory entropy, a causal edge is established between the dissolved oxygen concentration in the water body and the fish movement trajectory entropy. For example, if the dissolved oxygen offset in a certain area decreases by 1.3 and the fish movement trajectory entropy in the same area increases by 0.9 at the same time, a causal edge is established between the dissolved oxygen concentration in the water body and the fish movement trajectory entropy, and a causal weight of 0.7 is set.

[0037] Step S33, construct the rice-fish data heterogeneous graph based on the heterogeneous nodes and the rules for the edge relationships of the heterogeneous graph.

[0038] Specifically, traverse all rice grid cells, water areas, and sensor positions to obtain multiple rice nodes, fish nodes, and environmental nodes. Calculate the geometric distances between the rice nodes, fish nodes, and environmental nodes to obtain initial edges, and generate bidirectional edges, competitive edges, association edges, and causal edges based on the heterogeneous graph edge relationship rules. Construct a rice-fish data heterogeneous graph based on the rice nodes, fish nodes, environmental nodes, and heterogeneous graph edges.

[0039] Step S4: Construct a rice-fish attention network model. Import the rice-fish data heterogeneous graph into the rice-fish attention network model, and perform cross-domain feature fusion on the rice-fish symbiotic ecological data based on the rice-fish attention network model to obtain rice-fish association data.

[0040] In this embodiment, step S4 includes: Step S41: Construct a rice-fish attention network model.

[0041] In this embodiment, step S41 includes: Step S41-1: When constructing the rice-fish attention network model, import a dynamic meta-learning framework into the rice-fish attention network model.

[0042] Step S41-2: Divide the rice-fish symbiotic area into block partitions, divide the rice-fish symbiotic area into equal-area rectangular sub-regions, and obtain the historical rice-fish symbiotic ecological data of each rectangular sub-region.

[0043] Step S41-3: Use the historical rice-fish symbiotic ecological data of the same rectangular sub-region in different periods as training positive samples, randomly combine the historical rice-fish symbiotic ecological data of different rectangular sub-regions in the same period, and use the randomly combined historical rice-fish symbiotic ecological data as training negative samples.

[0044] Step S41-4: Train the rice-fish attention network model based on the training positive samples and training negative samples.

[0045] Step S42: Import the preprocessed rice-fish symbiotic ecological data and the rice-fish data heterogeneous graph into the rice-fish attention network model, obtain the feature vectors of the rice-fish symbiotic ecological data based on the rice-fish attention network model, and perform spatial projection on the feature vectors of the rice-fish symbiotic ecological data to map the feature vectors of the rice-fish symbiotic ecological data to the same space.

[0046] Specifically, based on the heterogeneous graph of rice-fish data, the heterogeneous node attributes of rice, fish, and the environment are obtained, and the heterogeneous node attributes of rice, fish, and the environment are integrated into corresponding node feature matrices. For example, the node attributes of rice node A01 are [chlorophyll = 0.65, nitrogen accumulation = 0.44, coordinates (15982, 35746)], the node attributes of rice node A02 are [chlorophyll = 0.58, nitrogen accumulation = 0.43, coordinates (15982, 35750)], and the node attributes of rice node A03 are [chlorophyll = 0.74, nitrogen accumulation = 0.41, coordinates (15982, 35754)]. Then the rice node feature matrix composed of these three rice nodes is: ; Each row vector in this matrix represents a rice node, and the column vectors represent chlorophyll content, canopy nitrogen accumulation, and the spatial coordinates of the rice node from left to right.

[0047] For example, the node attributes of fish node B31 are [density = 2.7, feeding = 13 times / 10 minutes, trajectory entropy = 1.7, coordinates (13785, 24658)], the node attributes of fish node B32 are [density = 2.5, feeding = 14 times / 10 minutes, trajectory entropy = 1.8, coordinates (13795, 24658)], and the node attributes of fish node B33 are [density = 2.1, feeding = 15 times / 10 minutes, trajectory entropy = 1.6, coordinates (13805, 24658)]. Then the fish node feature matrix composed of these three fish nodes is: ; Each row vector in this matrix represents a fish node, and the column vectors represent fish population density, fish feeding frequency, fish movement trajectory entropy, and spatial coordinates from left to right.

[0048] For example, the node attributes of environment node C11 are [dissolved oxygen concentration offset = -1.1, ammonia nitrogen concentration = 0.8, water temperature = 0.6, pH = 0.4], the node attributes of environment node C12 are [dissolved oxygen concentration offset = -1.2, ammonia nitrogen concentration = 0.7, water temperature = 0.6, pH = 0.4], and the node attributes of environment node C13 are [dissolved oxygen concentration offset = -1.3, ammonia nitrogen concentration = 0.8, water temperature = 0.5, pH = 0.4]. Then the environment node feature matrix composed of these three environment nodes is: ; Each row vector in this matrix represents an environment node, and the column vectors represent dissolved oxygen offset, ammonia nitrogen concentration, water temperature, and pH value from left to right.

[0049] Furthermore, construct an adjacency matrix to store the heterogeneous graph edge relationships existing between nodes. The heterogeneous graph edge relationships include bidirectional edges, symbiotic edges, competitive edges, and causal edges. For example, there is a bidirectional edge connection between rice node A01 and environmental node C11, a symbiotic edge connection between fish node B32 and environmental node C12, a competitive edge connection between environmental node C11 and rice node A03, and a causal edge connection between environmental node C13 and fish node B31. Then, the adjacency matrix storing the above edge connections is as follows: ; Each row vector of this matrix represents a heterogeneous graph edge relationship. The column vectors represent the source node, target node, edge relationship type, and weight from left to right. Among them, the edge relationship type uses to represent a bidirectional edge connection, to represent a symbiotic edge connection, to represent a competitive edge connection, and causal edge connection.

[0050] Step S43: Obtain the ecological attention weights of the nodes included in the rice-fish data heterogeneous graph, perform rice-fish association discovery on the rice-fish symbiotic ecological data feature vectors based on the ecological attention weights, and obtain rice-fish association data.

[0051] Specifically, obtain the source nodes and target nodes with edge connections in the rice-fish data heterogeneous graph, calculate the ecological attention weights between the source nodes and target nodes according to the ecological attention weight formula, and construct an ecological attention weight matrix. The ecological attention weight formula is expressed as ; where is the ecological attention weight, is the bidirectional edge connection weight between the source node and the target node, is the symbiotic edge connection weight between the source node and the target node, is the competitive edge connection weight between the source node and the target node, is the causal edge connection weight between the source node and the target node.

[0052] For example, there are bidirectional edge connection, symbiotic edge connection, competitive edge connection, and causal edge connection between rice node A01 and environmental node C11. The bidirectional edge connection weight is 0.7, the symbiotic edge connection weight is 0.5, the competitive edge connection weight is 0.2, and the causal edge connection weight is 0.6. Then, the ecological attention weight between rice node A01 and environmental node C11 is 0.52.

[0053] Further, perform correlation strength screening on the ecological attention weight matrix to obtain rice-fish correlation data. The correlation strength screening is expressed as dividing the data with ecological attention weights in the ecological attention weight matrix within [0.6, 1] into high-level correlation data, dividing the data with ecological attention weights in the range of [0.3, 0.6) into medium-level correlation data, and dividing the data with ecological attention weights in the range of [0, 0.3) into low-level correlation data.

[0054] Step S5: Based on the rice-fish correlation data, perform rice-fish causal reasoning to generate a rice-fish association knowledge graph.

[0055] In this embodiment, step S5 includes: Step S51: Perform optimized PC calculation on the rice-fish correlation data, and during the optimized PC calculation, add time lag constraints and ecological prior constraints to obtain the linear causal relationship between rice and fish based on the optimized PC calculation.

[0056] Specifically, first, generate lagged observation variables of the rice-fish correlation data based on the ecological response characteristics. The lag order is determined by maximizing the mutual information to ensure that the cause variables in the current causal reasoning only come from historical data. For example, align the dissolved oxygen concentration and water temperature in the environmental data with a 2 - 3-hour lag to the fish population density. Second, enforce the causal direction that historical data determines current data, and prohibit future variables from affecting past variables. For example, allow causal reasoning where a change in historical water temperature data leads to a change in fish population density, and prohibit causal reasoning where a change in fish population density leads to a change in water temperature data. Also, in the conditional independence test, limit the conditional set to only include earlier variables to avoid time series inversion. Finally, use the greedy algorithm to dynamically optimize the lag order of each variable. If the mutual information of a certain lag order with the biological indicator is significantly higher than other orders, fix this order to capture the maximum causal effect.

[0057] Further, delete the causal relationships in the rice-fish correlation data that violate biological common sense. For example, delete the causal relationship in the rice-fish correlation data where the change in fish population density causes the change in water temperature data. Skip the independence test and retain the known strong correlation causal relationships. For example, forcibly retain the strong correlation causal relationship that the dissolved oxygen concentration in the water body affects the fish population density. Limit the range of test variables. For example, when testing the correlation between water temperature and rice chlorophyll content, forcibly add the dissolved oxygen concentration data in the water body to the correlation test process between water temperature and rice chlorophyll content. The optimized PC algorithm accurately extracts the linear causal relationship between rice and fish by integrating time lag constraints and ecological prior constraints.

[0058] Step S52: Perform non-linear extension on the non-Gaussian distributed data in the rice-fish correlation data to identify and obtain the non-linear causal relationships in the rice-fish correlation data.

[0059] Specifically, for the non-Gaussian distributed data in the rice-fish association data, complex causal relationships are captured through data preprocessing and non-linear testing methods. First, non-Gaussian variables in the rice-fish association data are screened by the KS test, and polynomial terms and interaction terms are introduced to capture the non-linear interaction effects of the data in the rice-fish association data. Subsequently, non-linear causal relationships in the rice-fish association data are obtained with the assistance of deep learning.

[0060] Step S53: Generate an initial rice-fish causal network based on the non-linear causal relationship and the linear causal relationship.

[0061] Step S54: Calculate the rice-fish treatment effect data of rice and fish, and construct a rice-fish association knowledge graph based on the initial rice-fish causal network and the rice-fish treatment effect data of rice and fish.

[0062] Step S6: Establish a rice-fish decision model, import the rice-fish association knowledge graph into the rice-fish decision model, and generate a rice-fish regulation decision. The user can dynamically regulate the rice-fish symbiotic area based on the rice-fish regulation decision.

[0063] Specifically, fuse the real-time rice-fish symbiotic ecological data with the rice-fish association knowledge graph to construct a rice-fish prediction matrix. For example: ; The above matrix is the rice-fish prediction matrix of a certain rice-fish symbiotic area, where each row vector represents the rice-fish symbiotic ecological data of the rice-fish symbiotic area within a certain time period, and the column vectors represent the time period, water temperature, oxygen content concentration, ammonia nitrogen content concentration, and fish feeding frequency from left to right in sequence. The time period includes the current time period t, the historical time periods t - 2h and t - 1h50m, that is, the data two hours ago and the data one hour and fifty minutes ago, and the predicted time periods t + 10m and t + 2h, that is, the predicted data ten minutes later and the predicted data two hours later.

[0064] Furthermore, the data in the rice-fish association knowledge graph is divided into an environmental dimension, a biological dimension, and an operation dimension. The environmental dimension, biological dimension, and operation dimension are combined into a decision space. The environmental dimension includes two dimension features: a safety threshold range and a risk level. The biological dimension includes two dimension features: the rice growth stage and the fish activity state. The operation dimension includes two dimension features: direct regulation actions and indirect regulation strategies. For example, in a certain rice-fish symbiotic area, the safety threshold range of the environmental dimension includes a water temperature safety threshold range of [25, 30°C], a dissolved oxygen concentration safety threshold range of [3, 5 mg / L], and an ammonia nitrogen content concentration safety threshold range of [0.4, 0.8 mg / L]. The risk level dimension feature is divided into three risk levels: low, medium, and high according to the degree of deviation from the safety threshold. In the biological dimension, the rice growth stage is divided into three growth stages: tillering stage, heading stage, and maturity stage by the chlorophyll content SPAD. When SPAD < 30, it is the tillering stage; when 30 ≤ SPAD < 40, it is the heading stage; when SPAD ≥ 40, it is the maturity stage. The fish activity state is divided into three states: normal, warning, and escape according to the fish density Den per unit area. When Den ≥ 80 tails / mu, it is the normal state; when 50 tails / mu ≤ Den < 80 tails / mu, it is the warning state; when Den < 50 tails / mu, it is the escape state. The direct regulation actions of the operation dimension include oxygenation, water change, feeding adjustment, and fertilization. The indirect regulation strategies include fry density adjustment and rice variety replacement.

[0065] Generate decisions based on the causal relationships contained in the rice-fish association knowledge graph. Set the immediate execution type of decisions as explicit regulation decisions and the long-term execution type of decisions as implicit regulation decisions to obtain a set of regulation decisions. For example, there is a causal edge in the rice-fish association knowledge graph: "The decrease in fish density is caused by the dissolved oxygen concentration being lower than 2.8 mg / L", then this causal edge is generated into a regulation decision: "If the dissolved oxygen concentration is lower than 2.8 mg / L and the duration exceeds two hours, start the aerator for oxygenation."

[0066] Construct a state-decision mapping table based on the set of regulation decisions. The row vectors of the state-decision mapping table are represented as combinations of rice-fish ecological states. For example, state 1 is {water temperature = 34°C, dissolved oxygen concentration = 2.8 mg / L, SPAD = 28}, and state 2 is {ammonia nitrogen concentration = 0.9 mg / L, Den = 65 tails / mu, SPAD = 44}. The column vectors of the state-decision mapping table are represented as candidate regulation decisions. For example, {decision: start the aerator, trigger condition: dissolved oxygen concentration < 2.8 mg / L or water temperature > 30°C and dissolved oxygen concentration < 4.0 mg / L, intensity: running duration = (target dissolved oxygen concentration - current dissolved oxygen concentration) / 0.5}.

[0067] Establish regulatory decision priority rules. For example, set the implementation priority of explicit regulatory decisions to be greater than that of implicit regulatory decisions, the implementation priority of decision-making in high-risk state areas to be greater than that in medium- and low-risk state areas, give priority to processing decisions with higher absolute weights under the same risk level, and the priority of non-linear causal relationships to be greater than that of linear causal relationships.

[0068] Construct a rice-fish decision model based on the decision space, state-decision mapping table, and regulatory decision priority rules. Users can obtain regulatory decisions for the rice-fish symbiotic area based on the rice-fish decision model and precisely regulate the rice-fish symbiotic area.

[0069] Figure 2 The schematic diagram of a smart agriculture management system based on image processing provided by some embodiments of the present application, which can implement the idea of the present application, is shown. Figure 3 The schematic diagram of the rice-fish symbiotic area network in a smart agriculture management system based on image processing provided by some embodiments of the present application, which can implement the idea of the present application, is shown. The following provides a detailed introduction to this smart agriculture management system based on image processing.

[0070] Specifically, a smart agriculture management system based on image processing includes: A rice-fish symbiotic area network, which is used to collect rice data, fish data, and environmental data in the rice-fish symbiotic area to obtain rice-fish symbiotic ecological data.

[0071] A composition module, which is used to perform data preprocessing on the rice-fish symbiotic ecological data and construct a heterogeneous graph of rice-fish data.

[0072] A rice-fish model module, which is used to construct a rice-fish attention network model and obtain rice-fish association data based on the heterogeneous graph of rice-fish data and the rice-fish symbiotic ecological data.

[0073] A knowledge construction module, which is used to perform rice-fish causal reasoning and generate a rice-fish association knowledge graph.

[0074] A rice-fish decision module, which is used to generate rice-fish regulatory decisions. Users can dynamically regulate the rice-fish symbiotic area based on the rice-fish regulatory decisions.

[0075] Furthermore, the rice-fish symbiotic area network includes: A rice data collection terminal, which includes a drone cluster. The drone cluster is used to cruise the rice-fish symbiotic area with a cruise frequency of twice a day, and obtain chlorophyll content and canopy nitrogen accumulation based on the drone cluster.

[0076] Fish data acquisition terminal, the fish data acquisition terminal includes an underwater sonar array and an infrared module, and monitors fish based on the underwater sonar array and the infrared module to obtain fish movement trajectories, fish feeding frequencies, and fish population densities.

[0077] Environmental data acquisition terminal, the environmental acquisition terminal is used to monitor the environment of the rice-fish symbiotic area to obtain the dissolved oxygen concentration of the water body, the nitrogen ammonia concentration of the water body, the pH value of the water body, and the water temperature.

[0078] The specific usage method and function of this embodiment will be described below: First, deploy a network for the rice-fish symbiotic area, and collect multi-source data on rice, fish, and the environment based on the network of the rice-fish symbiotic area to obtain rice-fish symbiotic ecological data. Then, perform data preprocessing on the rice-fish symbiotic ecological data to eliminate the influence of abnormal data on the rice-fish symbiotic ecological data and improve the accuracy of the rice-fish symbiotic ecological data. Based on the preprocessed rice-fish symbiotic ecological data, construct a rice-fish data heterogeneous graph. Subsequently, construct a rice-fish attention network model, import the rice-fish data heterogeneous graph into the rice-fish attention network model, and perform cross-domain feature fusion on the rice-fish symbiotic ecological data based on the rice-fish attention network model to obtain rice-fish association data. Then, perform rice-fish causal reasoning based on the rice-fish association data to generate a rice-fish association knowledge graph. Finally, establish a rice-fish decision model, import the rice-fish association knowledge graph into the rice-fish decision model, and generate rice-fish regulation decisions. In this embodiment, by constructing a rice-fish data heterogeneous graph and a rice-fish association knowledge graph, multi-source data and complex ecological relationships in the rice-fish symbiotic area are integrated, and the complex ecological relationships in the rice-fish symbiotic area are quantified into a specific set of causal relationships. Combining the set of causal relationships with real-time data provides users with more accurate and personalized intelligent dynamic regulation decisions.

[0079] In addition, an embodiment of the present invention also provides an electronic device, including: At least one processor; and a memory communicatively connected to at least one processor; wherein, the memory stores instructions executable by at least one processor, and the instructions are executed by at least one processor so that at least one processor can execute the method proposed in Embodiment 1 of the present invention.

[0080] The following specifically introduces each component of the electronic device: Among them, the processor is the control center of the electronic device, which can be a single processor or a collective term for multiple processing elements. For example, the processor can be one or more central processing units (CPUs), or can be an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement Embodiment 1 of the present invention. For example, one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0081] Among them, the processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0082] The memory is used to store the software program for implementing the solution of the present invention and is controlled by the processor for execution. The specific implementation manner can refer to the above method embodiments and will not be elaborated here.

[0083] The memory can be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or can also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory can be integrated with the processor or exist independently and be coupled to the processor through the interface circuit of the electronic device. The embodiments of the present invention do not make specific limitations on this.

[0084] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by means of a limited (such as infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that contains one or more sets of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be solid-state drives.

[0085] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context before and after.

[0086] It should be understood that in the embodiments of the present invention, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0087] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A smart agricultural management method based on image processing, characterized in that: The method comprises: Deploy a rice-fish symbiosis area network for a rice-fish symbiosis area, wherein the rice-fish symbiosis area is a rice-fish symbiosis area where fish are cultivated while rice is planted; Based on the rice-fish symbiosis regional network, multi-source data of rice, fish and environment are collected to obtain rice-fish symbiosis ecological data, wherein the rice-fish symbiosis ecological data includes rice data, fish data and environmental data, and data preprocessing is performed on the rice-fish symbiosis ecological data; Construct a rice-fish data heterogeneous graph based on the rice-fish symbiotic ecological data after data preprocessing; Construct a rice-fish attention network model, import the heterogeneous graph of rice-fish data into the rice-fish attention network model, perform cross-domain feature fusion of rice-fish symbiotic ecological data based on the rice-fish attention network model, and obtain rice-fish related data; Conduct rice-fish causal reasoning based on rice-fish association data and generate a rice-fish association knowledge graph; A rice-fish decision-making model is established, the rice-fish associated knowledge graph is imported into the rice-fish decision-making model, and a rice-fish regulation decision is generated. Users can dynamically regulate the rice-fish symbiotic area based on the rice-fish regulation decision.

2. According to claim 1, a smart agricultural management method based on image processing is characterized in that: The method of importing the rice-fish data heterogeneous graph into the rice-fish attention network model, performing cross-domain feature fusion on the rice-fish symbiotic ecological data based on the rice-fish attention network model, and obtaining rice-fish associated data includes: The preprocessed rice-fish symbiotic ecological data and the rice-fish data heterogeneous graph are imported into the rice-fish attention network model, the feature vector of the rice-fish symbiotic ecological data is obtained based on the rice-fish attention network model, and the feature vector of the rice-fish symbiotic ecological data is spatially projected to map the feature vector of the rice-fish symbiotic ecological data to the same space; The ecological attention weights of the nodes contained in the heterogeneous graph of rice-fish data are obtained, and the rice-fish association is discovered on the characteristic vectors of rice-fish symbiotic ecological data based on the ecological attention weights to obtain rice-fish association data.

3. According to claim 1, a smart agricultural management method based on image processing is characterized in that: The construction of the rice-fish attention network model includes: When constructing a rice-fish attention network model, a dynamic meta-learning framework is introduced into the rice-fish attention network model; Divide the rice-fish symbiosis area into blocks, divide the rice-fish symbiosis area into rectangular sub-areas of equal area, and obtain the historical rice-fish symbiosis ecological data of each rectangular sub-area; The historical rice-fish symbiosis ecological data of different periods in the same rectangular sub-region are used as training positive samples, and the historical rice-fish symbiosis ecological data of the same period in different rectangular sub-regions are randomly combined, and the historical rice-fish symbiosis ecological data after random combination are used as training negative samples; The rice-fish attention network model is trained based on training positive samples and training negative samples.

4. The method for intelligent agricultural management based on image processing according to claim 1, characterized in that: The method of performing rice-fish causal reasoning based on the rice-fish association data and the rice-fish data heterogeneous graph to generate a rice-fish association knowledge graph includes: Optimizing PC calculations on rice-fish association data, and adding time lag constraints and ecological prior constraints in the process of optimizing PC calculations, wherein the time lag constraints are used to constrain the causes of causal reasoning, and the ecological prior constraints are used to constrain the reasoning logic of causal reasoning, obtaining the linear causal relationship between rice and fish based on the optimized PC calculations, performing nonlinear expansion on non-Gaussian distributed data in the rice-fish association data, identifying and obtaining the nonlinear causal relationship in the rice-fish association data, and generating an initial rice-fish causal network based on the nonlinear causal relationship and the linear causal relationship; The rice-fish treatment effect data of rice and fish are calculated, and a rice-fish association knowledge graph is constructed based on the initial rice-fish causal network and the rice-fish treatment effect data of rice and fish.

5. The method for intelligent agricultural management based on image processing according to claim 1, characterized in that: The step of establishing a rice-fish decision model, importing a rice-fish related knowledge graph into the rice-fish decision model, and generating a rice-fish regulation decision includes: A rice-fish decision space is constructed based on the rice-fish association knowledge graph and combined with rice-fish symbiotic ecological data. Rice-fish ecological prediction is performed according to the initial rice-fish causal network contained in the rice-fish association knowledge graph. Rice-fish regulatory decisions are generated based on the results of rice-fish ecological prediction. The rice-fish regulatory decisions include explicit regulatory decisions and implicit regulatory decisions.

6. The method for intelligent agricultural management based on image processing according to claim 5 is characterized in that: The rice-fish decision space is constructed based on the rice-fish association knowledge graph and combined with the rice-fish symbiotic ecological data, including: Obtain rice knowledge nodes, fish knowledge nodes and environmental knowledge nodes in the rice-fish association knowledge graph, obtain knowledge node data based on the rice knowledge nodes, fish knowledge nodes and environmental knowledge nodes, convert the knowledge node data into a rice-fish decision space vector, and convert the original rice-fish causal network contained in the rice-fish association knowledge graph into a rice-fish association weight matrix; The rice-fish decision space is constructed based on the rice-fish decision space vector and the rice-fish association weight matrix.

7. The method for intelligent agricultural management based on image processing according to claim 1, characterized in that: The method of constructing a rice-fish data heterogeneous graph based on the rice-fish symbiotic ecological data after data preprocessing includes: The preprocessed rice-fish symbiotic ecological data is used to define graph nodes, obtain rice heterogeneous nodes, fish heterogeneous nodes and environmental heterogeneous nodes, add edge relationship rules to the rice heterogeneous nodes, fish heterogeneous nodes and environmental heterogeneous nodes, and link the rice heterogeneous nodes, fish heterogeneous nodes and environmental heterogeneous nodes based on the edge relationship rules to generate a heterogeneous graph of rice-fish data.

8. The method for intelligent agricultural management based on image processing according to claim 1, characterized in that: The method collects multi-source data of rice, fish and environment based on the rice-fish symbiosis regional network to obtain rice-fish symbiosis ecological data, wherein the rice-fish symbiosis ecological data includes rice data, fish data and environmental data, including: The rice data collection terminal, fish data collection terminal and environmental data collection terminal based on the rice-fish symbiosis regional network obtain rice-fish symbiosis ecological data, the rice data includes chlorophyll content and canopy nitrogen accumulation, the fish data includes fish movement trajectory, fish feeding frequency and fish population density, and the environmental data includes water dissolved oxygen concentration, water nitrogen and ammonia concentration, water pH value and water temperature.

9. A smart agricultural management system based on image processing, characterized in that: include: A rice-fish symbiosis regional network, wherein the rice-fish symbiosis regional network is used to collect rice data, fish data and environmental data in the rice-fish symbiosis region to obtain rice-fish symbiosis ecological data; A mapping module, which is used to pre-process the rice-fish symbiotic ecological data and construct a heterogeneous graph of the rice-fish data; A rice-fish model module, wherein the rice-fish model module is used to construct a rice-fish attention network model and obtain rice-fish association data based on a rice-fish data heterogeneous graph and rice-fish symbiotic ecological data; A knowledge construction module, wherein the knowledge construction module is used to perform rice-fish causal reasoning and generate a rice-fish association knowledge graph; The rice-fish decision module is used to generate rice-fish regulation decisions, and users can dynamically regulate the rice-fish symbiotic area based on the rice-fish regulation decisions.

10. The intelligent agricultural management system based on image processing according to claim 9, characterized in that: The rice-fish symbiosis regional network includes: A rice data collection terminal, comprising a drone cluster, which is used to patrol the rice-fish symbiosis area twice a day, and obtain chlorophyll content and canopy nitrogen accumulation based on the drone cluster; A fish data collection terminal, which includes an underwater sonar array and an infrared module, monitors fish based on the underwater sonar array and the infrared module to obtain fish movement trajectories, fish feeding frequencies, and fish population densities; The environmental data acquisition terminal is used to monitor the environment of the rice-fish symbiosis area and obtain the dissolved oxygen concentration, nitrogen and ammonia concentration, pH value and temperature of the water.

Images