Watermelon planting environment monitoring and disease and pest prevention and control system

Abstract

The invention discloses a watermelon planting environment monitoring and pest prevention and control system, which comprises a weather acquisition integration module, a water quality and water source monitoring module, a soil moisture content monitoring module, a real-time weather forecast module, an irrigation control module, a real-time detection analysis module, a data integration display module, a decision support module, a monitoring early warning module and an optimization adjustment module, through the decision support module, farmers can be helped to recognize possible problems in the growth process in advance, the irrigation amount and the fertilization opportunity are optimized, the use efficiency of resources is improved, human decision errors are reduced, and the precision is improved; meanwhile, an irrigation strategy can be dynamically adjusted through the irrigation control module according to real-time data and weather forecast, excessive irrigation or water shortage is avoided, and therefore the utilization efficiency of water resources is remarkably improved, water waste is reduced, it is ensured that water supply in each stage meets the actual requirements of watermelons, healthy growth of the watermelons is promoted, and the economic benefit is increased. And unnecessary irrigation caused by meteorological changes is effectively avoided, so that the operation cost is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of agricultural planting technology, and particularly relates to a watermelon planting environment monitoring and pest control system. Background Art

[0002] In recent years, with the intensification of global climate change and the pressure of population growth, agricultural production has faced unprecedented challenges. How to improve agricultural production efficiency and crop quality under limited land and resource conditions has become an important issue in the global agricultural field. Especially in the aspect of efficient and refined crop management, intelligent technologies based on big data have brought new opportunities to agriculture. The wide application of big data technology, especially in agricultural planting, can not only monitor various environmental parameters in real time, but also provide accurate decision-making support for agricultural production through data analysis and modeling. Watermelon, as a crop sensitive to the growth environment, its yield and quality are often affected by multiple environmental factors such as temperature, humidity, precipitation, and soil pH value. In the traditional watermelon planting management mode, it often relies on manual experience for judgment, which is easily limited by the space and time of crop growth. Therefore, how to obtain relevant data in a timely manner through effective environmental monitoring means and make scientific and reasonable planting decisions based on these data has become an important research direction for improving watermelon planting efficiency and quality;

[0003] The existing watermelon environment monitoring system cannot identify potential problems that may occur during the growth process in advance, reducing the efficiency of resource use and having a high error rate in human decision-making. In addition, the existing watermelon environment monitoring system cannot dynamically adjust the irrigation strategy according to real-time data and weather forecasts, easily resulting in over-irrigation or water shortage, and cannot avoid unnecessary irrigation caused by meteorological changes, increasing the operating cost. For this reason, we propose a watermelon environment monitoring system based on big data. Summary of the Invention

[0004] The purpose of the present invention is to provide a watermelon planting environment monitoring and pest control system for the problems that the existing watermelon environment monitoring system cannot identify potential problems that may occur during the growth process in advance, reducing the efficiency of resource use and having a high error rate in human decision-making;

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: A watermelon planting environment monitoring and pest control system, including: a meteorological collection integration module, a water quality and water source monitoring module, a soil moisture monitoring module, a real-time weather forecast module, an irrigation control module, a real-time detection and analysis module, a data integration and display module, a decision support module, a monitoring and early warning module, an optimization and adjustment module, a data security module, and a user management module;

[0006] The meteorological collection integration module, water quality and water source monitoring module, soil moisture monitoring module, and real-time weather forecast module are used to obtain real-time monitoring information;

[0007] The irrigation control module adjusts and controls the irrigation system according to the input real-time monitoring information, and then inputs the irrigation plan into the decision support module, optimization adjustment module, and data integration display module respectively;

[0008] The real-time detection and analysis module collects melon field data and pest and disease information in real time, and generates pest control suggestions based on the input real-time monitoring information and the input data of the optimization adjustment module;

[0009] The decision support module predicts the growth status of watermelons under different environmental conditions based on historical watermelon growth data, the input real-time monitoring information, and the input pest and disease information, and provides planting decisions;

[0010] The optimization adjustment module adjusts the allocation of agricultural resources according to the input real-time monitoring information, irrigation plan, pest and disease information, and planting decision, and feeds back the adjusted agricultural resource allocation information to the decision support module and the real-time detection and analysis module;

[0011] The monitoring and early warning module analyzes the pest and disease information input by the real-time detection and analysis module in real time and automatically generates early warning information;

[0012] The data integration display module processes and integrates the input real-time monitoring data, planting decisions, agricultural resource allocation, and early warning information, and displays the real-time status of the melon field environment through a visual interface. Moreover, it stores all information in the data security module and is connected to the user management module.

[0013] Furthermore, the real-time monitoring information includes real-time meteorological data, water quality data, soil data, the predicted value of the watermelon growth status corresponding to the current soil data, and weather forecast data;

[0014] Among them, the real-time meteorological data is obtained by the meteorological collection integration module, the water quality data is obtained by the water quality and water source monitoring module, the soil data and the predicted value of the watermelon growth status corresponding to the current soil data are obtained by the soil moisture monitoring module, and the weather forecast data is obtained by the real-time weather forecast module.

[0015] Furthermore, the specific steps for the soil moisture monitoring module to obtain the predicted value of the watermelon growth status corresponding to the current soil data are as follows:

[0016] S1.1: The soil moisture monitoring module collects various soil data collected by each sensor, preprocesses each set of collected data, then scales each item of data to the range of [0, 1] through normalization processing. After that, the soil data at different times after processing is used as a feature vector, and each set of processed soil data is divided into a training set, a test set, and a validation set;

[0017] S1.2: Based on the Bi-GRU model architecture, a relationship analysis model is established. Then the training set is input into the relationship analysis model. The relationship analysis model performs forward propagation on the training set data at each moment, processes the forward hidden state and the backward hidden state of the input data through the forward GRU and the backward GRU respectively. After that, the relationship analysis model concatenates the hidden states of the forward and backward GRUs to obtain the final output, and then performs non-linear processing on the final output through the output layer, and generates the predicted value of the watermelon growth state for output;

[0018] S1.3: Calculate the loss value between the predicted value and the actual value through the mean square error function, input the loss value from the output layer of the relationship analysis model, and transfer the loss value layer by layer based on the chain rule. Calculate the gradient of this loss value for each layer of the relationship analysis model through the backpropagation algorithm, and then use the gradient descent algorithm to optimize the parameters of each layer of the relationship analysis model;

[0019] S1.4: After each round of training is completed, input the validation set data into the relationship analysis model. The relationship analysis model calculates the predicted value of each validation set data through forward propagation, and calculates the loss value between the predicted value and the actual value to evaluate the performance indicators of this relationship analysis model. If the performance indicators reach the preset performance indicators, stop training; otherwise, repeat training and validation until the preset number of iterations is reached;

[0020] S1.5: Use the test set to verify the prediction accuracy of the trained relationship analysis model on unknown data, and deploy the verified relationship analysis model to the environmental monitoring platform. Then input the real-time soil data into the relationship analysis model, and start from the first time step through the forward GRU of the relationship analysis model to process the soil data at each moment in turn. The backward GRU starts from the last time step and processes the data in reverse order. After that, use the output of the bidirectional GRU as the input and pass it to the output layer, and generate the predicted value of the watermelon growth state corresponding to the current soil data after post-processing by the output layer, and then input it into the irrigation control module, the real-time detection and analysis module, the decision support module, the data integration and display module, the monitoring and early warning module, and the data security module.

[0021] Further, the specific steps for the irrigation control module to adjust and control the irrigation system are as follows:

[0022] S2.1: The irrigation control module receives the pre - processed real - time meteorological data, water quality data, soil data, and weather prediction data, constructs a corresponding objective function based on the real - time meteorological data, weather prediction data, water quality data, and soil data. Then, it initializes multiple groups of irrigation control schemes based on historical scheme data and expert experience, generates a population according to the generated irrigation schemes, where the individuals in the population represent a group of irrigation control schemes respectively. The combination of decision variables in each irrigation control scheme is used as a path;

[0023] S2.2: Initialize the pheromone values of all paths to a unified constant, calculate the objective function values of each group of individuals, and take the negative value of its objective function value as the heuristic value. Then, calculate the probability of selecting the next optional decision variable under the current decision variable according to the pheromone value and the heuristic value, and select the decision variable with the highest probability value ranked first from large to small based on the probabilities of each group of decision variables;

[0024] S2.3: After the construction of each group of individual paths is completed, that is, a complete irrigation control scheme is generated. Calculate the objective function value of the generated irrigation control scheme, and update the pheromone value on the path according to the objective function value. Then, re - perform path selection and pheromone value update until the objective function values of each group of irrigation control schemes converge within the preset threshold;

[0025] S2.4: Compare the objective function values of each group of irrigation control schemes, and select the scheme with an objective function value higher than the preset threshold as the recommended irrigation control scheme for the staff to view and select, and implement it in the actual melon field irrigation. At the same time, randomly perturb the irrigation control scheme to generate multiple groups of control schemes as a new population for the next round of scheme update.

[0026] Furthermore, the specific steps for the real - time detection and analysis module to generate pest control suggestions are as follows:

[0027] S3.1: Collect various groups of pest and disease information through various sensors and cameras arranged in the melon field, fill in the missing data in each group of pest and disease information, and then extract feature data from the original pest and disease information. Based on the real - time monitored meteorological data, weather prediction data, water quality data, soil data, current agricultural resource configuration, and pest and disease growth laws, simulate the spread of pests and diseases through the SIR model, and evaluate the current spread speed and infection degree of pests and diseases;

[0028] S3.2: After generating the pest and disease risk assessment results, use the current melon field state as the root node, including the real - time monitoring data and environmental characteristics of pests and diseases. Extract the pest control measures that can be taken in the current melon field from the pest control measure database, and use the melon field state after the implementation of each pest control measure as the child node, and use the pest control measure as the edge to connect the child node with its corresponding parent node to construct a corresponding tree structure;

[0029] S3.3: Starting from the root node, i.e., the current melon field state, calculate the upper confidence bound values of each child node. Through the UCB selection strategy, layer by layer, select the child node with the highest upper confidence bound value among the child nodes until an unvisited and not fully developed leaf node is reached. Then, generate new child nodes through the pest control measures available at this node, add them to the tree structure, and perform pest control simulation based on this node. Stop the simulation until the pest control reaches the target. At the same time, backtrack the simulation results to each node on this path, and update the reward values and visit counts of each node;

[0030] S3.4: Repeat the selection, expansion, simulation, and backtracking until the preset search depth is reached. Traverse the finally generated tree structure, and layer by layer, select the nodes with reward values higher than the preset threshold to generate a complete pest control strategy, including the type of pest control measures, the time of pest control measures, and the evaluation of pest control effects. And display this pest control strategy through the data integration display model for the melon field management personnel to view.

[0031] Further, the decision support module predicts the growth status of watermelons under different environmental conditions, and the specific steps for providing planting decisions are as follows:

[0032] S4.1: Collect meteorological data, water quality data, soil data, weather forecasts, irrigation plans, watermelon pest and disease data, and the predicted values of the growth status of watermelons corresponding to the current soil data through sensors and monitoring devices. Preprocess each group of collected data, and construct a hidden state set S = {S1, S2, S3, S4, S5} based on the predicted values of the growth status of watermelons corresponding to the current soil data, where S1 represents the germination period of watermelons, S2 represents the seedling period of watermelons, S3 represents the flowering period of watermelons, S4 represents the fruiting period of watermelons, and S5 represents the maturity period of watermelons. Then, take the environmental data such as soil humidity, soil temperature, air temperature, precipitation, and meteorological data as observation variables, and construct the corresponding observation variable set;

[0033] S4.2: Based on historical watermelon growth data, calculate the transition frequency of watermelons from one growth state to another, and normalize the calculation results to obtain the transition probability. Then, according to the relationship between the growth state of watermelons and environmental data, establish an observation probability distribution model corresponding to each hidden state, and calculate the conditional probability of the observed data under each hidden state based on historical environmental data and the actual growth state of watermelons to train the observation probability distribution;

[0034] S4.3: Input the real-time collected environmental data into the trained observation probability distribution model, infer the current hidden state of watermelons through Bayesian filtering, update the hidden state probability distribution at the current moment according to the inference result, and at the same time generate the growth prediction of the watermelons in this melon field and generate corresponding planting decision suggestions.

[0035] Further, the specific steps for the optimization and adjustment module to adjust the agricultural resource allocation according to the input real-time monitoring information, irrigation plan, pest and disease information, and planting decision are as follows:

[0036] S5.1: Initialize a population containing multiple groups of agricultural resource allocation plans based on real-time meteorological data, water quality data, soil data, predicted values of watermelon growth status corresponding to the current soil data, weather prediction data, real-time planting decision, irrigation plan, and watermelon pest and disease data. Each individual represents a configuration plan. Calculate the fitness value of each individual's current agricultural resource allocation plan based on the current watermelon growth situation, and traverse the fitness values of each individual in the initial population. Select the individual with the highest fitness value ranked from large to small as the head solution of the current population;

[0037] S5.2: Set a group of coefficient vectors. At the same time, in each iteration, generate a random number between 0 and 1. If the generated random number is greater than the preset probability parameter, calculate the coefficient A and step size L for controlling the position update according to the coefficient vector, and judge whether the other individuals shrink and surround around the head solution according to the magnitude of the direction;

[0038] S5.3: If |A| < 1, it means that the other individuals surround around the head solution, and update the positions of the other individuals according to the head solution, that is, adjust the resource allocation in the agricultural resource allocation plan of each individual. If |A| ≥ 1, it means that the other individuals are far away from the head solution. Through random search, select a random position in the population space to update the positions of the other individuals, and update the coefficient vector based on the linearly decreasing rule after each iteration;

[0039] S5.4: If the generated random number is less than or equal to the preset probability parameter, calculate the distance between the positions of the other individuals in the population and the position of the head solution, and simulate the movement law of each individual approaching the head solution along the spiral trajectory through the spiral motion formula to update the individual positions;

[0040] S5.5: Repeatedly perform head solution selection and iterative update of positions until the fitness value converges within the preset threshold range. Then compare the fitness values in each group of individuals, and output the agricultural resource allocation plan with the highest fitness value ranked from large to small to obtain the adjusted agricultural resource allocation information, and adjust the resource allocation situation of the current melon fields based on the adjusted agricultural resource allocation information.

[0041] Beneficial effects:

[0042] 1. The present invention models the hidden state of the watermelon growth status, and through the integrated analysis of soil data, meteorological data, and watermelon growth status, it can help farmers identify potential problems in the growth process in advance. At the same time, by predicting the growth stage and status changes of watermelons, combined with environmental data, it optimizes the irrigation volume and fertilization timing. Moreover, through the optimization adjustment module, it automatically evaluates the usage of various resources, reasonably allocates the input amount of each resource, improves the resource utilization efficiency, reduces human decision-making errors, and enhances the accuracy.

[0043] 2. The present invention initializes multiple groups of irrigation control schemes based on historical scheme data and expert experience, and generates a population according to the generated irrigation schemes. The individuals in the population respectively represent a group of irrigation control schemes. The combination of decision variables in each irrigation control scheme is used as a path. The pheromone values of all paths are initialized to a unified constant, and the objective function values of each group of individuals are calculated, and the negative value of its objective function value is used as the heuristic value. Then, according to the pheromone value and the heuristic value, the probability of selecting the next optional decision variable under the current decision variable is calculated. Based on the probability values of each group of decision variables, the decision variable ranked first in descending order of probability values is selected. After the path of each group of individuals is constructed, a complete irrigation control scheme is generated. The objective function value of the generated irrigation control scheme is calculated, and the pheromone value on the path is updated according to the objective function value. Then, the path selection and pheromone value update are carried out again until the objective function values of each group of irrigation control schemes converge within the preset threshold. The objective function values of each group of irrigation control schemes are compared, and the scheme with an objective function value higher than the preset threshold is selected as the recommended irrigation control scheme for the staff to view and select and implement in the actual watermelon field irrigation. At the same time, the irrigation control scheme is randomly perturbed to generate multiple groups of control schemes as a new population for the next round of scheme update. It can dynamically adjust the irrigation strategy according to real-time data and weather forecasts, avoid over-irrigation or water shortage, thereby significantly improving the utilization efficiency of water resources, reducing water waste, and ensuring that the water supply in each stage meets the actual needs of watermelons, promoting their healthy growth, effectively avoiding unnecessary irrigation caused by meteorological changes, and thus reducing the operating cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a system framework diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0045] The following further explains the present invention with reference to the accompanying drawings.

[0046] Such as Figure 1As shown in the figure, the present invention provides a watermelon planting environment monitoring and pest control system, which is characterized by including: a meteorological data acquisition and integration module, a water quality and water source monitoring module, a soil moisture monitoring module, a real-time weather forecast module, an irrigation control module, a real-time detection and analysis module, a data integration and display module, a decision support module, a monitoring and early warning module, an optimization and adjustment module, a data security module, and a user management module;

[0047] The meteorological data acquisition and integration module is used to collect meteorological data and analyze it to obtain the change trend of the current weather on the same day, that is, to obtain real-time meteorological data and input it into the irrigation control module, the real-time detection and analysis module, the decision support module, the optimization and adjustment module, and the data integration and display module.

[0048] The water quality and water source monitoring module is used to collect various water quality index data of the water source in real time, that is, to obtain water quality data and input it into the irrigation control module, the real-time detection and analysis module, the decision support module, the optimization and adjustment module, and the data integration and display module.

[0049] The real-time weather forecast module is used to obtain meteorological forecast data within a preset future time period, including air quality, temperature change, and rainfall prediction, that is, to obtain weather prediction data and input it into the irrigation control module, the real-time detection and analysis module, the decision support module, the optimization and adjustment module, the data integration and display module, the monitoring and early warning module, and the data security module.

[0050] The soil moisture monitoring module uses sensors to monitor multi-layer soil data and analyze the relationship between soil conditions and the growth status of watermelons, that is, to obtain soil data and the predicted value of the growth status of watermelons corresponding to the current soil data, and input the monitored and analyzed data into the irrigation control module, the real-time detection and analysis module, the decision support module, the data integration and display module, the monitoring and early warning module, and the data security module respectively.

[0051] The irrigation control module is used to adjust and control the irrigation system according to the input data of the meteorological data acquisition and integration module, the water quality and water source monitoring module, the real-time weather forecast module, and the soil moisture monitoring module, and then input the irrigation plan into the decision support module, the optimization and adjustment module, and the data integration and display module respectively.

[0052] The real-time detection and analysis module is used to collect melon field data and pest information in real time, and generate pest control suggestions based on the input data of the meteorological data acquisition and integration module, the real-time weather forecast module, the water quality and water source monitoring module, the soil moisture monitoring module, and the optimization and adjustment module, and then input it into the decision support module, the optimization and adjustment module, the data integration and display module, and the monitoring and early warning module.

[0053] The decision support module predicts the growth status of watermelons under different environmental conditions based on historical watermelon growth data, input real-time monitoring information, and input pest and disease information, provides planting decisions, and then inputs them into the optimization and adjustment module.

[0054] The optimization and adjustment module adjusts the allocation of agricultural resources according to the input real-time monitoring information, irrigation plan, pest and disease information, and planting decisions, and feeds back the adjusted agricultural resource allocation information to the decision support module and the real-time detection and analysis module.

[0055] The monitoring and early warning module is used to analyze the pest and disease information input by the real-time detection and analysis module in real time and automatically generate early warning information.

[0056] The data integration and display module is used to process and integrate the input real-time monitoring data, planting decisions, agricultural resource allocation, and early warning information, display the real-time status of the melon field environment through a visual interface, store all information in the data security module, and connect to the user management module. The data security module is used to manage the data of the entire system, and the user management module is used to manage the user permissions of the entire system.

[0057] In the soil moisture monitoring module, the specific steps for the soil moisture monitoring module to obtain the predicted value of the watermelon growth status corresponding to the current soil data are as follows:

[0058] S1.1: The soil moisture monitoring module collects various soil data collected by each sensor, preprocesses the collected data sets, scales each data through normalization processing to the range of [0, 1], then takes the processed soil data at different times as a feature vector, and divides the processed soil data sets into a training set, a test set, and a validation set.

[0059] S1.2: Based on the Bi-GRU model architecture, a relationship analysis model is established, and then the training set is input into the relationship analysis model. The relationship analysis model performs forward propagation on the training set data at each moment, processes the forward hidden state and the reverse hidden state of the input data through the forward GRU and the reverse GRU respectively, then the relationship analysis model concatenates the hidden states of the forward and reverse GRUs to obtain the final output, and then performs non-linear processing on the final output through the output layer and generates the predicted value output of the watermelon growth status.

[0060] S1.3: Calculate the loss value between the predicted value and the actual value through the mean square error function, input the loss value from the output layer of the relationship analysis model, and layer by layer transfer the loss value based on the chain rule, calculate the gradient of the loss value for each layer of the relationship analysis model through the backpropagation algorithm, and then use the gradient descent algorithm to optimize the parameters of each layer of the relationship analysis model.

[0061] S1.4: After each round of training, input the validation set data into the relationship analysis model. The relationship analysis model calculates the predicted values of each validation set data through forward propagation, and calculates the loss value between the predicted values and the actual values to evaluate various performance indicators of the relationship analysis model. If the performance indicators reach the preset performance indicators, stop training; otherwise, repeat training and validation until the preset number of iterations is reached.

[0062] S1.5: Use the test set to verify the prediction accuracy of the trained relationship analysis model on unknown data, and deploy the verified relationship analysis model to the environmental monitoring platform. Then input the real-time soil data into the relationship analysis model. Starting from the first time step, the forward GRU in the relationship analysis model processes the soil data at each moment in sequence, and the reverse GRU starts from the last time step and processes the data in reverse order. Then, the output of the bidirectional GRU is used as the input and passed to the output layer. After post-processing by the output layer, the predicted value of the watermelon growth state corresponding to the current soil data is generated, and then it is input into the irrigation control module, real-time detection and analysis module, decision support module, data integration and display module, monitoring and early warning module, and data security module.

[0063] In the irrigation control module, the specific steps for the irrigation control module to adjust and control the irrigation system are as follows:

[0064] S2.1: The irrigation control module receives the preprocessed real-time meteorological data, water quality data, soil data, and weather prediction data, constructs the corresponding objective function based on the real-time meteorological data, weather prediction data, water quality data, and soil data. Then, initialize multiple groups of irrigation control schemes based on historical scheme data and expert experience, and generate a population according to the generated irrigation schemes. The individuals in the population represent a group of irrigation control schemes respectively. Combine the decision variables in each irrigation control scheme as a path.

[0065] S2.2: Initialize the pheromone values of all paths to a unified constant, calculate the objective function values of each group of individuals, and use the negative value of their objective function values as the heuristic value. Then, calculate the probability of selecting the next optional decision variable under the current decision variable based on the pheromone value and the heuristic value, and select the decision variable with the largest probability value ranked first based on the probability values of each group of decision variables.

[0066] S2.3: After the construction of each group of individual paths is completed, that is, a complete irrigation control scheme is generated. Calculate the objective function value of the generated irrigation control scheme, and update the pheromone value on the path according to the objective function value. Then, re-select the path and update the pheromone value until the objective function values of each group of irrigation control schemes converge within the preset threshold.

[0067] S2.4: Compare the objective function values of each group of irrigation control schemes, and select the scheme with an objective function value higher than the preset threshold as the recommended irrigation control scheme for the staff to view and select, and implement it in the actual melon field irrigation. At the same time, randomly perturb the irrigation control scheme to generate multiple groups of control schemes as a new population for the next round of scheme update.

[0068] In the real-time detection and analysis module, the specific steps for generating pest control suggestions by the real-time detection and analysis module are as follows:

[0069] S3.1: Collect the pest and disease information of each group through various sensors and cameras arranged in the melon field, fill in the missing data in each group of pest and disease information, and then extract the feature data from the original pest and disease information. Based on the real-time monitored meteorological data, weather prediction data, water quality data, soil data, current agricultural resource allocation, and the growth law of pests and diseases, simulate the spread of pests and diseases through the SIR model, and evaluate the current spread speed and infection degree of pests and diseases.

[0070] S3.2: After generating the pest and disease risk assessment results, use the current melon field state as the root node, including the real-time monitoring data and environmental characteristics of pests and diseases. Extract the pest control measures that can be taken for the current melon field from the pest control measure database, and use the melon field state after the implementation of each pest control measure as the child node, and use the pest control measure as the edge to connect the child node with its corresponding parent node to construct the corresponding tree structure.

[0071] S3.3: Starting from the root node, that is, the current melon field state, calculate the upper confidence bound value of each child node, and select the child node with the highest upper confidence bound value layer by layer downward through the UCB selection strategy until an unvisited and not fully developed leaf node is reached. Then, generate new child nodes through the pest control measures that can be taken by this node and add them to the tree structure, and perform pest control simulation based on this node until the pest control reaches the target and the simulation stops. At the same time, trace back the simulation results to each node on this path, and update the reward value and access times of each node.

[0072] S3.4: Repeat the selection, expansion, simulation, and backtracking until the preset search depth is reached, traverse the finally generated tree structure, and select the nodes with reward values higher than the preset threshold layer by layer to generate a complete pest control strategy, including the type of pest control measures, the time of pest control measures, and the evaluation of pest control effects, and display the pest control strategy through the data integration display model for the melon field management staff to view.

[0073] In the decision support module, the specific steps for the decision support module to predict the growth status of watermelons under different environmental conditions and provide planting decisions are as follows:

[0074] S4.1: Collect meteorological data, water quality data, soil data, as well as weather forecasts, irrigation plans, watermelon pest and disease data, and the predicted values of the watermelon growth state corresponding to the current soil data through sensors and monitoring devices. Preprocess each group of collected data. Based on the predicted values of the watermelon growth state corresponding to the current soil data, construct a hidden state set S = {S1, S2, S3, S4, S5}, where S1 represents the germination period of the watermelon, S2 represents the seedling period of the watermelon, S3 represents the flowering period of the watermelon, S4 represents the fruiting period of the watermelon, and S5 represents the maturity period of the watermelon. Then, take soil humidity, soil temperature, air temperature, precipitation, and each environmental data of meteorological data as observation variables, and construct the corresponding observation variable set.

[0075] S4.2: Based on historical watermelon growth data, calculate the transition frequency of the watermelon from one growth state to another, and perform normalization processing on the calculation results to obtain transition probabilities. Then, according to the relationship between the growth state of the watermelon and environmental data, establish an observation probability distribution model corresponding to each hidden state, and calculate the conditional probability of the observed data under each hidden state based on historical environmental data and the actual growth state of the watermelon to train the observation probability distribution.

[0076] S4.3: Input the real-time collected environmental data into the trained observation probability distribution model, infer the current hidden state of the watermelon through Bayesian filtering, update the hidden state probability distribution at the current moment according to the inference result, and simultaneously generate the growth prediction of the watermelon in this melon field and generate corresponding planting decision suggestions.

[0077] In the optimization and adjustment module, the specific steps for the optimization and adjustment module to adjust agricultural resource allocation according to the input real-time monitoring information, irrigation plan, pest and disease information, and planting decision are as follows:

[0078] S5.1: According to the real-time meteorological data, water quality data, soil data, the predicted values of the watermelon growth state corresponding to the current soil data, weather prediction data, real-time planting decision, irrigation plan, and watermelon pest and disease data, initialize a population containing multiple groups of agricultural resource allocation plans, where each individual represents a configuration plan. Calculate the fitness value of each individual's current agricultural resource allocation plan based on the current watermelon growth situation, and traverse the fitness values of each individual in the initial population. Select the individual with the largest fitness value ranked first and use it as the head solution of the current population.

[0079] S5.2: Set a group of coefficient vectors. At the same time, in each round of iteration, generate a random number between 0 and 1. If the generated random number is greater than the preset probability parameter, calculate the coefficient A and step size L for controlling position update according to the coefficient vector, and judge whether the remaining individuals shrink and surround around the head solution according to the magnitude of the direction.

[0080] S5.3: If |A| < 1, it means that the remaining individuals surround the head solution, and the positions of the remaining individuals are updated according to the head solution, that is, the resource allocation in the agricultural resource allocation scheme of each individual is adjusted. If |A| ≥ 1, it means that the remaining individuals move away from the head solution. Through random search, a random position is selected in the population space, and the positions of the remaining individuals are updated. After each iteration ends, the coefficient vector is updated based on the rule of linear decrease;

[0081] S5.4: If the generated random number is less than or equal to the preset probability parameter, calculate the distance between the positions of the remaining individuals in the population and the position of the head solution, and simulate the movement law of each individual approaching the head solution along the spiral trajectory through the spiral motion formula to update the individual position;

[0082] S5.5: Repeatedly perform the selection of the head solution and the iterative update of the position until the fitness value converges within the preset threshold range. Then, compare the fitness values in each group of individuals, and output the agricultural resource allocation scheme ranked first in descending order of fitness value to obtain the adjusted agricultural resource allocation information, and adjust the resource allocation situation of each current melon field based on the adjusted agricultural resource allocation information.

[0083] The present invention models the hidden state of the watermelon growth state through the decision support module, and with the help of the integrated analysis of soil data, meteorological data, and watermelon growth state, can help farmers identify potential problems in the growth process in advance. At the same time, by predicting the growth stage and state changes of watermelons, combined with environmental data, the irrigation amount and fertilization timing are optimized. And through the optimization adjustment module, the usage of various resources is automatically evaluated, the input amount of each resource is reasonably allocated, the usage efficiency of resources is improved, the human decision-making error is reduced, and the accuracy is improved.

[0084] The present invention also dynamically adjusts the irrigation strategy through the irrigation control module according to real-time meteorological data, water quality data, soil data, and weather forecasts, avoiding over-irrigation or water shortage, thereby significantly improving the utilization efficiency of water resources, reducing water waste, and ensuring that the water supply in each stage meets the actual needs of watermelons, promoting their healthy growth, effectively avoiding unnecessary irrigation caused by meteorological changes, thereby reducing the operation cost. At the same time, the real-time detection and analysis module is used to generate pest control strategies by collecting pest information of each group in real time, so as to understand the pest situation in the melon field in real time and control it in real time, further improving the watermelon yield and quality.

[0085] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A watermelon planting environment monitoring and pest control system, characterized in that Including: A meteorological collection integration module, a water quality and water source monitoring module, a soil moisture monitoring module, a real-time weather forecast module, an irrigation control module, a real-time detection and analysis module, a data integration and display module, a decision support module, a monitoring and early warning module, an optimization and adjustment module, a data security module, and a user management module; The meteorological collection integration module, the water quality and water source monitoring module, the soil moisture monitoring module, and the real-time weather forecast module are used to obtain real-time monitoring information; The irrigation control module adjusts and controls the irrigation system according to the input real-time monitoring information, and then inputs the irrigation plan into the decision support module, the optimization and adjustment module, and the data integration and display module respectively; The real-time detection and analysis module collects melon field data and pest and disease information in real time, and generates pest control suggestions based on the input real-time monitoring information and the input data of the optimization and adjustment module; The decision support module predicts the growth status of watermelons under different environmental conditions based on historical watermelon growth data, the input real-time monitoring information, and the input pest and disease information, and provides planting decisions; The optimization and adjustment module adjusts the allocation of agricultural resources according to the input real-time monitoring information, irrigation plan, pest and disease information, and planting decision, and feeds back the adjusted agricultural resource allocation information to the decision support module and the real-time detection and analysis module; The monitoring and early warning module analyzes the pest and disease information input by the real-time detection and analysis module in real time and automatically generates early warning information; The data integration and display module processes and integrates the input real-time monitoring data, planting decisions, agricultural resource allocation, and early warning information, and displays the real-time status of the melon field environment through a visual interface. Moreover, it stores all information in the data security module and is connected to the user management module.

2. The watermelon planting environment monitoring and pest control system according to claim 1, characterized in that, The real-time monitoring information includes real-time meteorological data, water quality data, soil data, the predicted value of the watermelon growth status corresponding to the current soil data, and weather forecast data; Among them, the real-time meteorological data is obtained by the meteorological collection integration module, the water quality data is obtained by the water quality and water source monitoring module, the soil data and the predicted value of the watermelon growth status corresponding to the current soil data are obtained by the soil moisture monitoring module, and the weather forecast data is obtained by the real-time weather forecast module.

3. The watermelon planting environment monitoring and pest control system according to claim 2, wherein The specific steps for the soil moisture monitoring module to obtain the predicted value of the watermelon growth status corresponding to the current soil data are as follows: S1.1: The soil moisture monitoring module collects various soil data collected by each sensor, preprocesses the collected data groups, then scales each item of data to the range of [0, 1] through normalization processing. After that, the processed soil data at different times is used as a feature vector, and the processed soil data groups are divided into a training set, a test set, and a validation set; S1.2: Based on the Bi-GRU model architecture, establish a relationship analysis model. Then input the training set into the relationship analysis model. The relationship analysis model performs forward propagation on the training set data at each moment, processes the forward hidden state and backward hidden state of the input data through the forward GRU and backward GRU respectively. After that, the relationship analysis model concatenates the hidden states of the forward and backward GRUs to obtain the final output. Then, the output layer performs non-linear processing on the final output and generates the predicted value of the watermelon growth state for output; S1.3: Calculate the loss value between the predicted value and the actual value through the mean square error function. Input the loss value from the output layer of the relationship analysis model, and layer by layer transfer the loss value based on the chain rule. Calculate the gradient of this loss value for each layer of the relationship analysis model through the backpropagation algorithm, and then use the gradient descent algorithm to optimize the parameters of each layer of the relationship analysis model; S1.4: After each round of training, input the validation set data into the relationship analysis model. The relationship analysis model calculates the predicted values of each validation set data through forward propagation and calculates the loss value between the predicted value and the actual value to evaluate the performance indicators of the relationship analysis model. If the performance indicators reach the preset performance indicators, stop training; otherwise, repeat training and validation until the preset number of iterations is reached; S1.5: Use the test set to verify the prediction accuracy of the trained relationship analysis model on unknown data, and deploy the verified relationship analysis model to the environmental monitoring platform. Then input the real-time soil data into the relationship analysis model. Starting from the first time step, the forward GRU of the relationship analysis model processes the soil data at each moment in sequence, and the backward GRU starts from the last time step and processes the data in reverse order. After that, take the output of the bidirectional GRU as the input and pass it to the output layer. After post-processing by the output layer, generate the predicted value of the watermelon growth state corresponding to the current soil data, and then input it into the irrigation control module, real-time detection and analysis module, decision support module, data integration and display module, monitoring and early warning module, and data security module.

4. The watermelon planting environment monitoring and pest control system according to claim 2, characterized in that, The specific steps for the irrigation control module to adjust and control the irrigation system are as follows: S2.1: The irrigation control module receives the preprocessed real-time meteorological data, water quality data, soil data, and weather prediction data, constructs the corresponding objective function based on the real-time meteorological data, weather prediction data, water quality data, and soil data. Then initialize multiple groups of irrigation control schemes based on historical scheme data and expert experience, and generate a population according to the generated irrigation schemes. The individuals in the population represent a group of irrigation control schemes respectively. Combine the decision variables in each irrigation control scheme as a path; S2.2: Initialize the pheromone values of all paths to a unified constant, calculate the objective function values of each group of individuals, and take the negative value of their objective function values as the heuristic value. Then calculate the probability of selecting the next optional decision variable under the current decision variable based on the pheromone value and the heuristic value, and select the decision variable with the largest probability value ranked first based on the probabilities of each group of decision variables; S2.3: After the construction of each group of individual paths is completed, a complete irrigation control plan is generated, the objective function value of the generated irrigation control plan is calculated, and the pheromone value on the path is updated according to the objective function value. Then, path selection and pheromone value update are performed again until the objective function values of each group of irrigation control plans converge within a preset threshold; S2.4: Compare the objective function values of each group of irrigation control plans, and select the plan with an objective function value higher than the preset threshold as the recommended irrigation control plan for the staff to view and select and implement in the actual melon field irrigation. At the same time, randomly perturb the irrigation control plan to generate multiple groups of control plans as a new population for the next round of plan update.

5. The watermelon planting environment monitoring and pest control system according to claim 2, wherein, The specific steps for the real-time detection and analysis module to generate pest control suggestions are as follows: S3.1: Collect pest and disease information for each group through various sensors and cameras arranged in the melon field, fill in the missing data in each group of pest and disease information, and then extract feature data from the original pest and disease information. Based on the real-time monitored meteorological data, weather prediction data, water quality data, soil data, current agricultural resource allocation, and pest and disease growth laws, simulate the spread of pests and diseases through the SIR model, and evaluate the current spread speed and infection degree of pests and diseases; S3.2: After generating the pest and disease risk assessment results, use the current melon field state as the root node, including the real-time monitoring data and environmental characteristics of pests and diseases, extract the pest control measures that can be taken in the current melon field from the pest control measure database, and use the melon field state after the implementation of each pest control measure as the child node, and use the pest control measure as the edge to connect the child node to its corresponding parent node to construct the corresponding tree structure; S3.3: Starting from the root node, that is, the current melon field state, calculate the upper confidence bound value of each child node, and select the child node with the highest upper confidence bound value layer by layer downward through the UCB selection strategy until an unvisited and not fully developed leaf node is reached. Then, generate new child nodes through the pest control measures that can be taken by this node and add them to the tree structure, and perform pest control simulation based on this node until the pest control reaches the target and the simulation stops. At the same time, trace back the simulation results to each node on this path, and update the reward value and visit times of each node; S3.4: Repeat the selection, expansion, simulation, and backtracking until the preset search depth is reached, traverse the finally generated tree structure, and select the nodes with reward values higher than the preset threshold layer by layer to generate a complete pest control strategy, including the type of pest control measures, the time of pest control measures, and the evaluation of pest control effects, and display the pest control strategy through the data integration display model for the melon field management staff to view.

6. The watermelon planting environment monitoring and pest control system according to claim 2, wherein The specific steps for the decision support module to predict the growth status of watermelons under different environmental conditions and provide planting decisions are as follows: S4.1: Collect meteorological data, water quality data, soil data, as well as weather forecasts, irrigation plans, watermelon pest and disease data, and predicted values of watermelon growth status corresponding to the current soil data through sensors and monitoring devices. Preprocess each set of collected data. Based on the predicted values of watermelon growth status corresponding to the current soil data, construct a hidden state set S = {S1, S2, S3, S4, S5}, where S1 represents the germination period of watermelon, S2 represents the seedling period of watermelon, S3 represents the flowering period of watermelon, S4 represents the fruiting period of watermelon, and S5 represents the maturity period of watermelon. Then, take soil humidity, soil temperature, air temperature, precipitation, and each environmental data of meteorological data as observation variables, and construct a corresponding observation variable set; S4.2: Based on historical watermelon growth data, calculate the transition frequency of watermelon from one growth state to another, and normalize the calculation results to obtain transition probabilities. Then, according to the relationship between the growth state of watermelon and environmental data, establish an observation probability distribution model corresponding to each hidden state, and calculate the conditional probability of observation data under each hidden state based on historical environmental data and the actual growth state of watermelon to train the observation probability distribution; S4.3: Input the real-time collected environmental data into the trained observation probability distribution model, speculate on the current hidden state of watermelon through Bayesian filtering, update the hidden state probability distribution at the current moment according to the speculation result, and simultaneously generate the growth prediction of the watermelon in this melon field and generate corresponding planting decision suggestions.

7. The watermelon planting environment monitoring and pest control system according to claim 6, wherein The specific steps for the optimization and adjustment module to adjust the allocation of agricultural resources according to the input real-time monitoring information, irrigation plan, pest and disease information, and planting decision are as follows: S5.1: Initialize a population containing multiple sets of agricultural resource allocation plans according to real-time meteorological data, water quality data, soil data, predicted values of watermelon growth status corresponding to the current soil data, weather forecast data, real-time planting decision, irrigation plan, and watermelon pest and disease data. Each individual represents an allocation plan. Calculate the fitness value of each individual's current agricultural resource allocation plan based on the current watermelon growth situation, and traverse the fitness values of each individual in the initial population. Select the individual with the highest fitness value ranked from large to small as the head solution of the current population; S5.2: Set a set of coefficient vectors. At the same time, in each iteration, generate a random number between 0 and 1. If the generated random number is greater than the preset probability parameter, calculate the coefficient A and step size L for controlling position update according to the coefficient vector, and judge whether the remaining individuals contract and surround around the head solution according to the magnitude of the direction; S5.3: If |A| < 1, it means that the remaining individuals surround around the head solution, and update the positions of the remaining individuals according to the head solution, that is, adjust the resource allocation in each individual's agricultural resource allocation plan. If |A| ≥ 1, it means that the remaining individuals move away from the head solution. Through random search, select a random position in the population space to update the positions of the remaining individuals, and update the coefficient vector based on the linearly decreasing rule after each iteration; S5.4: If the generated random number is less than or equal to the preset probability parameter, calculate the distances between the positions of the remaining individuals in the population and the position of the head solution, and simulate the movement law of each individual approaching the head solution along the spiral trajectory through the spiral movement formula to update the individual positions; S5.5: Repeatedly perform the head solution selection and iterative update of the positions until the fitness value converges within the preset threshold range. Then, compare the fitness values in each group of individuals, output the agricultural resource allocation plan ranked first in descending order of fitness value, obtain the adjusted agricultural resource allocation information, and adjust the resource allocation situation of each current melon field based on the adjusted agricultural resource allocation information.

Images