Fumigation control method based on pest and disease damage trend prediction

Abstract

The invention provides a fumigation control method based on pest and disease damage trend prediction, and the method comprises the steps: carrying out the spectrum analysis of audio data information through employing a fast Fourier transform algorithm, extracting the frequency domain feature vector of the chirp sound of each pest species, and enabling the frequency domain feature vector to comprise a fundamental frequency, harmonic energy distribution, chirp duration and chirp rhythm; according to the extracted chirp frequency domain feature vector, chirps of local main agricultural pests are identified and classified by adopting a support vector machine algorithm, and the types and life cycle information of the pests are judged by matching with a preset pest species chirp feature template library; and performing spatial mapping on the predicted population density distribution data by adopting a Kriging interpolation algorithm to generate a spatial distribution diagram, determining the aggregation degree and diffusion trend of pests in each region of the field, and dividing a plurality of prevention and control regions and buffer regions.

Description

Technical Field

[0001] The invention relates to the field of information technology, and in particular to a fumigation control method based on pest and disease trend prediction. Background Art

[0002] When using the chirping sounds of insects to monitor their physiological state and environmental adaptation, we are faced with a complex technical problem. There are many kinds of insects, and the chirping frequencies and intensities of different kinds of insects under different environmental conditions vary. The chirping characteristics of the same insect will also change significantly at different growth stages, in different activity states, and under different environmental conditions such as temperature, humidity, and food supply. In addition, in the natural environment, the chirping sounds of insects are usually mixed with background noises, such as wind, rain, and the calls of other animals. The frequency and intensity of these background noises will also change with time and environmental changes, posing a great challenge to the accurate extraction and identification of insect chirping sounds. At the same time, the chirping sounds of the same insect in different geographical locations and ecological environments may vary greatly. There may also be certain differences in the characteristics of insect chirping sounds; this requires the collection and analysis of a large number of insect chirping sound samples in different regions and environments, and the establishment of a sound feature library covering different situations; in practical applications, it is also necessary to consider how to collect and analyze insect chirping sounds in real time, continuously and on a large scale; this places high demands on the performance of monitoring equipment and the efficiency of data transmission, storage and processing; and, in harsh field environments, how to ensure the stability of monitoring equipment and the reliability of data is also a major challenge; in short, the technical difficulties faced in using insect chirping sounds for monitoring are multifaceted, involving multiple technical fields such as sound signal processing, pattern recognition, big data analysis, embedded systems, sensor networks, etc., which require in-depth exploration and research. Summary of the invention

[0003] The present invention provides a fumigation control method based on pest and disease trend prediction, which mainly includes:

[0004] The audio data of insect chirping sounds in a preset frequency band is obtained through a sound sensor, the audio data of insect chirping sounds is processed by a Butterworth filter, and the audio data of insect chirping sounds is decomposed by a wavelet transform method to obtain audio data information of chirping sounds of different insect species;

[0005] The fast Fourier transform algorithm is used to perform spectrum analysis on the audio data information to extract the frequency domain feature vector of each insect species' chirping sound. The frequency domain feature vector includes the fundamental frequency, harmonic energy distribution, chirping duration and chirping rhythm.

[0006] Based on the extracted frequency domain feature vectors of the chirping sounds, the support vector machine algorithm is used to identify and classify the chirping sounds of the main local agricultural pests. By matching them with the preset insect species chirping sound feature template library, the pest species and life cycle information are determined.

[0007] According to the information of pest species and life cycle, combined with the temperature and humidity data collected by the environmental temperature and humidity sensors, the logistic population growth model and Monte Carlo simulation algorithm are used to predict the growth curve of the pest population and the distribution data of insect population density. The distribution data of insect population density includes the mean value of insect population density and the confidence interval of insect population density. If the predicted mean value of insect population density exceeds the preset control threshold, a high-risk warning is output;

[0008] The Kriging interpolation algorithm is used to spatially map the predicted insect population density distribution data, generate a spatial distribution map, determine the concentration and diffusion trend of pests in various areas of the field, and divide several control areas and buffer zones;

[0009] For each control area, the particle swarm optimization algorithm is used to find the optimal fumigation plan between fumigation effect and cost, combining the area, pest species and insect population density distribution data of the control area. The optimal fumigation plan includes the combination of fumigant types, dosage ratio, fumigation time and temperature setting values.

[0010] The optimal fumigation scheme parameters for each control area are input into the automated fumigation robot control system. By adjusting the fumigant spraying amount, spraying duration and moving path, fixed-point, quantitative and timed fumigation operations are carried out in each control area. The fumigation process is monitored by an infrared thermal imager, and the fumigation intensity is dynamically adjusted according to real-time feedback.

[0011] During the fumigation process, audio data of pest chirping sounds in each prevention and control area is continuously collected, and the chirping sound intensity and frequency characteristic parameters are extracted. The fumigation effect is evaluated by comparing the chirping sound intensity and frequency characteristic parameters before and after fumigation. If the chirping sound parameters after fumigation do not reach the preset threshold, the secondary fumigation program is started. If the preset threshold is reached, the chirping sound intensity and frequency characteristic parameters before and after fumigation are uploaded to the cloud knowledge base.

[0012] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0013] The fumigation control method based on pest and disease trend prediction provided by the present invention can accurately extract and identify the chirping sounds of specific pests from complex environments through sound sensors and advanced audio processing technology, thereby improving the accuracy and speed of pest identification and helping to timely discover pest activities; combining the intelligent classification of the support vector machine algorithm with the population dynamics prediction of the logistic model, it can not only quickly identify the pest species, but also predict their population growth trends, achieve early warning of pest outbreaks, and provide a scientific basis for the formulation of prevention and control measures; using Kriging interpolation and Monte Carlo simulation, the pest distribution is presented in the form of a spatial map, providing intuitive pest density distribution information for farmland management, and dividing prevention and control measures. The system defines zones and buffer zones to guide precise application of pesticides, reduce excessive use of pesticides, and protect the ecological environment. Through the particle swarm optimization algorithm, the most suitable fumigation plan is customized for each prevention and control zone with comprehensive consideration of cost-effectiveness, thus achieving efficient allocation of resources and minimization of environmental impact. The application of automated fumigation robots, in conjunction with infrared thermal imaging monitoring, ensures the precise execution of fumigation operations and real-time feedback adjustment, improves operational efficiency and safety, and reduces manpower requirements. The comparative analysis of the chirping sound characteristics before and after fumigation provides instant feedback for evaluating the fumigation effect, ensuring the effectiveness of prevention and control measures. Through the construction of a cloud-based knowledge base, the system continuously accumulates and optimizes the experience and models of pest management, forming a closed-loop system of continuous improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a flow chart of a fumigation control method based on pest and disease trend prediction.

[0015] Figure 2 It is a schematic diagram of a fumigation control method based on pest and disease trend prediction of the present invention.

[0016] Figure 3 This is another schematic diagram of a fumigation control method based on pest and disease trend prediction according to the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figure 1-3 In this embodiment, a fumigation control method based on pest and disease trend prediction may specifically include:

[0019] Step S101, obtaining audio data of insect chirping sounds within a preset frequency band through a sound sensor, processing the audio data of insect chirping sounds using a Butterworth filter, and decomposing the audio data using a wavelet transform method to obtain audio data information of chirping sounds of different insect species.

[0020] A sound sensor is used to collect audio data of insect chirping sounds within a preset frequency band. The sampling rate and sampling accuracy are set according to the frequency characteristics of the chirping sounds of different insect species, the original audio data is collected, and the original audio data is converted into a digital signal; the digital signal is preprocessed, and the background noise existing in the original audio data is identified and marked through frequency domain analysis. According to the frequency distribution characteristics of the background noise, a Butterworth filter is used to effectively suppress and remove the background noise; the processed digital signal is decomposed in the time and frequency domain by a wavelet transform method, and the processed digital signal is decomposed into wavelet coefficients of different scales and frequencies; according to the energy distribution characteristics of the wavelet coefficients, the time and frequency domain characteristics of the chirping sounds of different insect species are judged to obtain the audio data information of the chirping sounds of different insect species.

[0021] Specifically, a sound sensor with a sensitivity of -38dB and a frequency response range of 20-150kHz is used to collect audio data of insect chirping sounds within a preset frequency band. According to the main frequency range of chirping sounds of common insect species such as katydids and crickets being 2-8kHz, the sampling rate is set to 41kHz and the sampling accuracy is set to 16bit. The original audio data is collected and converted into a digital signal through PCM encoding. The digital signal is preprocessed, and the time domain signal is converted to the frequency domain through FFT transformation. The low-frequency background noise components with large energy in the digital signal are identified, which are mainly distributed below 100Hz. Then, a fourth-order Butterworth high-pass filter is used, and the cutoff frequency is set to 100Hz to effectively suppress the background noise, and the noise reduction can reach more than 30dB. The fourth-order wavelet is used to decompose the filtered digital signal into 10 layers. Among the decomposed wavelet coefficients, the high-frequency components with larger energy are mainly concentrated in layers 3-6, corresponding to a frequency range of 2-8kHz, which is consistent with the chirping frequency of insect species such as katydids and crickets. Based on this, the energy threshold is set to extract the wavelet coefficients of this frequency band. The wavelet reconstruction method is used to obtain the final audio data of the chirping sounds of different insect species, thereby achieving effective separation and recognition of the chirping sounds.

[0022] Step S102, using a fast Fourier transform algorithm to perform spectrum analysis on the audio data information, extracting the frequency domain feature vector of the chirping sound of each insect species, the frequency domain feature vector including the fundamental frequency, harmonic energy distribution, chirping duration and chirping rhythm.

[0023] The audio data information of the chirping sounds of different insect species is obtained; the fast Fourier transform algorithm is used to perform spectrum analysis on the audio data to obtain a frequency domain feature vector, including fundamental frequency, harmonic energy distribution, chirping duration and chirping rhythm; the main frequency range of the chirping sound of each insect species is determined according to the fundamental frequency in the frequency domain feature vector; the spectrum structure and energy concentration area of ​​the chirping sound are judged by the harmonic energy distribution; the average duration and variation range of the chirping sound of each insect species is obtained according to the chirping duration; the chirping rhythm is used to analyze the time interval and regularity of the chirping sounds of the insect species; if the dimension of the frequency domain feature vector is greater than a preset threshold, the principal component analysis method is used to perform dimensionality reduction processing; the frequency domain feature vector is grouped by a clustering algorithm to obtain the category labels of the chirping sounds of different insect species.

[0024] Specifically, the collected audio data is first preprocessed, resampled with a sampling rate of 16kHz and a quantization bit number of 16 bits, and a bandpass filter is used to remove noise interference below 50Hz and above 8kHz. Then the audio data is framed using the Hamming window function, the frame length is set to 512 sampling points, and the frame shift is 256 sampling points to ensure that there is a 50% overlap between frames. A 512-point fast Fourier transform is performed on each frame of data to obtain the amplitude spectrum of 256 frequency components. By performing peak detection on the amplitude spectrum, the frequency component with the largest energy is obtained as the fundamental frequency, and the frequency range where the fundamental frequency is located is calculated. At the same time, several harmonic components with larger energy in the amplitude spectrum are counted to characterize the spectral structure characteristics of the chirping sound. By statistically analyzing the results of continuous multi-frame analysis, the average duration and rhythm characteristics of the chirping sound, such as the average number of chirping times per minute, are obtained. Finally, the features of each dimension are used to form a frequency domain feature vector, and the principal component analysis method is used to compress it to about 10 dimensions. The K-means clustering algorithm is used to cluster the frequency domain feature vectors of different insect species. The optimal number of clusters is obtained by calculating the Euclidean distance and silhouette coefficient between cluster centers, and a unique label is generated for each cluster.

[0025] Step S103, based on the extracted frequency domain feature vector of the chirping sound, a support vector machine algorithm is used to identify and classify the chirping sounds of the main local agricultural pests, and the type and life cycle information of the pests are determined by matching with a preset insect species chirping sound feature template library.

[0026] The extracted chirping sound frequency domain feature vectors are labeled, and a preset insect species chirping sound feature template library is constructed by analyzing the labeled chirping sound frequency domain feature vectors; the chirping sound frequency domain feature vectors are trained using a support vector machine algorithm, a chirping sound recognition and classification model is established, and the optimization is performed; the chirping sound frequency domain feature vectors collected in real time are input into the trained chirping sound recognition and classification model, and the type of pest is determined by matching with the preset insect species chirping sound feature template library; according to the determined pest type, combined with the preset pest life cycle feature information, the life cycle stage of the pest is analyzed and determined; if the identified pest type and the life cycle stage match the preset agricultural pest damage law, an early warning message is generated to indicate the risk of agricultural pest damage; the identification and classification results, pest life cycle information and early warning information are sent to agricultural production management personnel; the preset insect species chirping sound feature template library and the chirping sound recognition and classification model are continuously optimized and updated.

[0027] Specifically, the collected chirping sound is first preprocessed, and the time domain signal is converted into a frequency domain signal by fast Fourier transform, and the frequency domain feature vector is extracted. Then the extracted frequency domain feature vector is annotated, and the insect species corresponding to the feature vector is annotated. By statistically analyzing the annotated frequency domain feature vectors, a preset insect species chirping sound feature template library is constructed. The annotated frequency domain feature vectors are trained by a support vector machine algorithm, and the algorithm parameters are optimized by grid search and cross-validation methods, and finally a chirping sound recognition and classification model with an accuracy rate of more than 95% is established. The chirping sound frequency domain feature vector collected in real time is input into the trained model for prediction, and the pest species to which the chirping sound belongs is determined by calculating and matching the similarity with the feature vector in the preset insect species chirping sound feature template library. According to the determined pest species, combined with the morphological characteristics, activity patterns and other information of various pests in different life cycle stages pre-entered, reasoning analysis is performed through the life cycle model and knowledge base to determine the specific life cycle stage of the pest, such as egg stage, nymph stage, adult stage, etc. If the identified pest species and life cycle stages match the preset agricultural pest damage patterns, for example, when rice planthoppers appear in large numbers as adults, they will cause serious damage to rice, then an early warning message will be generated to indicate the possible risk of agricultural pest damage. The results of identification and classification, pest life cycle information, and early warning information will be sent to agricultural production management personnel in real time via text messages, APP push, etc., to provide a basis for decision-making in agricultural production. At the same time, the system will continue to optimize and update the preset insect species chirping feature template library and chirping recognition and classification model, and continuously supplement and improve the chirping data actually collected in various regions, increase the sample size, optimize algorithms and parameters, and improve the generalization ability of the model, so as to continuously improve the overall accuracy and reliability of agricultural pest chirping recognition and classification, and escort smart agricultural production.

[0028] Step S104, based on the pest species and life cycle information, combined with the temperature and humidity data collected by the environmental temperature and humidity sensor, the logistic population growth model and the Monte Carlo simulation algorithm are used to predict the growth curve of the pest population and the insect density distribution data; the insect density distribution data includes the insect density mean and the insect density confidence interval. If the predicted insect density mean exceeds the preset prevention and control threshold, a high-risk warning is output.

[0029] According to the pest species and life cycle information, a pest population growth model is established through the support vector machine algorithm to determine the key parameters of the pest population growth model; the temperature and humidity data collected by the environmental temperature and humidity sensor are obtained, and the temperature and humidity data are preprocessed and feature extracted to obtain the quantitative index of environmental factors; the quantitative index of environmental factors is used as input and substituted into the logistic population growth model to calculate the population growth rate and population size at the current moment; the Monte Carlo simulation algorithm is used to generate a random sample of population size according to the population size and the population growth rate; the random sample of population size is statistically analyzed to obtain the probability distribution of the population size, and the average insect population density is calculated. value and confidence interval of insect population density; determine whether the mean insect population density exceeds the preset prevention and control threshold. If the mean insect population density exceeds the preset prevention and control threshold, output high-risk warning information, including warning level and recommended prevention and control measures; if the mean insect population density does not exceed the preset prevention and control threshold, evaluate the pest risk level according to the confidence interval of the insect population density, and generate an insect monitoring plan; continuously monitor the changes in the quantitative indicators of the environmental factors and the population size, regularly update the population growth model and simulation prediction results, and optimize the preset prevention and control threshold and the insect monitoring plan; generate an insect density distribution map and a population growth curve through visualization technology to show the spatiotemporal dynamics of pest risks.

[0030] Specifically, firstly, according to the pest species and life cycle information, a pest population growth model is established through the support vector machine algorithm; the key parameters of the model include intrinsic growth rate r = 2, environmental capacity K = 1000, and initial population size N0 = 100; then the data collected by the environmental temperature and humidity sensor is obtained, the temperature and humidity data are denoised and normalized, and the two features of temperature and humidity are extracted, which are represented as T and H respectively. The environmental factors T and H are used as input and substituted into the logistic population growth model dN / dt = rN (1-N / K), and the population growth rate at the current moment is calculated to be 16 and the population size is 120. Then, the Monte Carlo simulation algorithm is used to generate 1000 random samples of population size based on the current population size of 120 and the population growth rate of 16. Statistical analysis of these 1000 samples shows that the population size follows a normal distribution with a mean of 126 and a standard deviation of 12, and the 95% confidence interval is calculated to be [102,150], and the mean insect density is 26 heads / square meter. If the mean insect density of 26 exceeds the preset control threshold of 0, a red high-risk warning message is output, and it is recommended to take chemical control measures immediately. If the mean insect density does not exceed the control threshold, the pest risk level in the next month is assessed as medium according to the confidence interval [102,150], and a monitoring plan is generated to monitor the population size once a week in the next month. The system will continuously monitor the dynamic changes of environmental temperature and humidity and population size, and use the latest monitoring data to update the population growth model and prediction results every half a month, and dynamically optimize the control threshold and monitoring frequency according to the prediction results. Finally, through visualization technology, a heat map of the spatiotemporal distribution of insect density and a population growth curve are generated to intuitively display the spatiotemporal dynamics of pest populations in the current and future periods, providing intuitive auxiliary decision support for agricultural management departments.

[0031] Step S105, using the Kriging interpolation algorithm to perform spatial mapping on the predicted insect population density distribution data, generate a spatial distribution map, determine the degree of pest aggregation and diffusion trend in each area of ​​the field, and divide a number of control areas and buffer zones.

[0032] The predicted insect population density distribution data is spatially mapped using the Kriging interpolation algorithm to obtain a spatial distribution map of the insect population density distribution; the degree of pest aggregation and the diffusion trend of each field area are determined based on the spatial distribution map of the insect population density distribution; the field area is clustered using a clustering algorithm based on the pest aggregation degree and the diffusion trend to divide a number of control areas and buffer zones; the geographical location information of each control area and buffer zone is obtained, and spatial positioning and boundary division are performed through a geographic information system; the control level and control measures of different areas are determined based on the pest aggregation degree and diffusion trend of each control area and buffer zone; if the pest aggregation degree of the control area is greater than the preset aggregation value and the diffusion trend is greater than If the concentration of pests in the control area is lower than the preset concentration value and the diffusion trend is less than the preset value, physical control measures are taken; according to the geographical location and area of ​​the buffer zone, a spatial interpolation algorithm is used to calculate the insect population density distribution information in the buffer zone to determine whether control measures need to be taken; if the insect population density in the buffer zone exceeds the preset threshold, the buffer zone is divided into control areas, and the control level and control measures are determined according to the insect population density; if the insect population density in the buffer zone does not exceed the preset threshold, the buffer zone is maintained unchanged; according to the divided control areas and buffer zones, a path planning algorithm is used to generate an agricultural machinery operation path to instruct the agricultural machinery to apply pesticides.

[0033] Specifically, when spatially mapping the predicted insect population density distribution data, the Kriging interpolation algorithm can be used. First, the field is divided into several 10m×10m grids, with the center point of the grid as the sample point. Then, the Kriging interpolation model is used to estimate the insect population density at any position in the grid according to the insect population density value of each sample point. The core of Kriging interpolation is to establish a semivariogram model. Commonly used semivariograms include exponential model, Gaussian model and spherical model. For example, using the exponential model, when the distance between sample points is h, the semivariogram is γ(h)=C0+C1(1-e^(-h / a)), where C0 is the nugget value, C1 is the base value, and a is the range. The semivariogram is fitted to the experimental variogram to obtain the model parameters, and then the insect population density of each grid is interpolated to generate an insect population density distribution map. According to the insect population density distribution map, the heat map visualization method is used to intuitively display the aggregation degree and diffusion trend of field pests. The insect population density is divided into 5 levels, represented by red, orange, yellow, green and blue respectively. The red area indicates the highest degree of pest aggregation, and the blue area indicates the lowest degree of pest aggregation. On this basis, the K-means clustering algorithm is used to divide the field into several control areas and buffer zones according to the insect population density value and spatial adjacency of each grid. The algorithm process is as follows: randomly select k grids as the initial cluster center; according to the Euclidean distance between each grid and the cluster center, divide it into the category where the nearest cluster center is located; according to the Euclidean distance between each grid and the cluster center, divide it into the category where the nearest cluster center is located; until the cluster center no longer changes. In the clustering results, grids with high insect population density and adjacent to each other are divided into control areas, and grids with low insect population density and located at the edge of the control area are divided into buffer zones. For the control area, when the insect population density exceeds 100 heads / square meter, chemical control measures are taken; when the insect population density is less than 100 heads / square meter, physical control measures are taken. For the buffer zone, the inverse distance weighted interpolation method is used to calculate its insect population density, with 50 insects / square meter as the threshold. Areas exceeding the threshold are divided into control areas and corresponding measures are taken. Finally, based on the division results, a genetic algorithm is used to optimize the operation path of agricultural machinery, with the center point of the control area as the path node and the distance between adjacent control areas as the weight, to search for the shortest path, guide agricultural machinery to accurately apply pesticides, and achieve differentiated control in different areas.

[0034] Step S106, for each divided control area, the particle swarm optimization algorithm is used to solve the optimal fumigation plan between the fumigation effect and the cost in combination with the area, pest species and insect population density distribution data of the control area. The optimal fumigation plan includes the combination of fumigant types, dosage ratio, fumigation time and temperature setting value.

[0035] The results of the control area division and the spatial distribution map of the pest density in each area are obtained, the area of ​​each control area is measured, and the geographical scope and area data of each control area are obtained; the pest species and insect population density distribution data of the divided control area are obtained, and the multi-objective optimization model is constructed by the particle swarm optimization algorithm in combination with the area data, wherein the objective function includes fumigation effect and cost; according to the pest species and the insect population density distribution data, the candidate fumigant type combination is determined to form an initial particle swarm, and each particle represents a fumigant type combination and a corresponding dosage ratio, fumigation time and temperature setting value; the fitness value of each particle is calculated, and the fitness value is combined with the two objective functions of the fumigation effect and cost, and is weighted according to the area of ​​the control area; the position and speed of the particle are updated, and according to the fitness value of the particle, the update formula of the particle swarm optimization algorithm is used, A new particle swarm is obtained; whether the new particle swarm satisfies the termination condition is determined, and if the termination condition is satisfied, the global optimal solution is output as the optimal fumigation scheme; if the termination condition is not satisfied, the position and velocity of the particle are updated, and a new particle swarm is obtained according to the fitness value of the particle and the update formula of the particle swarm optimization algorithm; according to the optimal fumigation scheme, the fumigant type combination, dosage ratio, fumigation time and temperature setting value of each prevention and control area are determined; environmental temperature and humidity data are obtained, and the fumigation time and temperature setting value are dynamically adjusted; by monitoring the change information of pest species and insect population density, the fumigation effect is evaluated, and the evaluation result is fed back to the multi-objective optimization model, and the parameters of the multi-objective optimization model are dynamically updated; according to the evaluation result of the fumigation effect and the cost accounting data, the multi-objective optimization model is used to continuously optimize the fumigation scheme and manage the fumigation process.

[0036] Specifically, first, the area data of the divided control areas are obtained through drone aerial photography and other methods. For example, the total area is 1,000 mu, of which area A is 200 mu, area B is 300 mu, and area C is 500 mu. At the same time, the pests are trapped and identified by traps, and the pest species and insect population density distribution in each area are determined. For example, area A is mainly locusts, with an average density of 50 heads / square meter; area B is mainly aphids, with an average density of 100 heads / square meter; area C is mainly borers, with an average density of 80 heads / square meter. Then, according to the pest species and insect population density, candidate fumigants such as cypermethrin and methyl bromide are determined, and parameters such as concentration, dosage, time, and temperature are set to form particles, and the initial particle group is constructed. The fitness value of each particle is calculated using the weighted comprehensive evaluation method, and the weight coefficient is determined according to the area of ​​each area, such as 2 in area A, 3 in area B, and 5 in area C. Then, the speed and position update formula of the particle swarm algorithm is used to obtain a new particle swarm. After 500 iterations, the termination condition is met and the optimal fumigation scheme is output: 50 ml / m3 of cypermethrin in area A, fumigation at 25℃ for 2 hours; 80 ml / m3 of methyl bromide in area B, fumigation at 30℃ for 3 hours; 60 ml / m3 of cypermethrin in area C, fumigation at 28℃ for 5 hours. During the fumigation process, the environmental parameters are obtained through the temperature and humidity sensors. If the temperature is 2℃ lower than the set value, the fumigation time is extended by 20%, and if it is 2℃ higher than the set value, the fumigation time is shortened by 20%, thereby dynamically adjusting the fumigation parameters. After fumigation, the types and numbers of pests in each area are regularly detected, and the fumigation effect is calculated, such as 95% in area A, 90% in area B, and 98% in area C. The weights of each parameter are dynamically updated by inputting the multi-objective optimization model. Finally, the fumigation effect and cost accounting data are combined to further optimize the next round of fumigation schemes and realize intelligent fumigation management.

[0037] Step S107, input the optimal fumigation scheme parameters of each control area into the automatic fumigation robot control system, and implement fixed-point, quantitative, and timed fumigation operations for each control area by adjusting the fumigant spraying amount, spraying time, and moving path. The fumigation process is monitored by an infrared thermal imager, and the fumigation intensity is dynamically adjusted according to real-time feedback.

[0038] The geographical location information and area data of each control area are obtained, and a grid division scheme for each control area is generated according to the area size and shape characteristics of the control area, and each control area is divided into a number of grid units; according to the degree of occurrence of pests and diseases and crop type information in each control area, the pest and disease control knowledge base is matched to obtain the optimal fumigation scheme parameters for each control area, and the optimal fumigation scheme parameter data is input into the automatic fumigation robot control system; according to the grid division scheme, a moving path of the automatic fumigation robot in each grid unit is generated, and the moving path needs to fully cover the area of ​​each grid unit; the automatic fumigation robot is controlled to navigate and move to the specified position of each grid unit according to the moving path, and the spraying position of the fumigant is determined according to the crop distribution map in the grid unit; after the automatic fumigation robot reaches the spraying position, the fumigant is sprayed according to the optimal fumigation scheme fumigant type and dosage parameters, adjust the fumigant spraying amount and spraying duration, and implement fumigation operations on crops in the grid unit; collect thermal imaging data of crop canopies in real time through infrared thermal imagers, analyze and calculate the temperature change trend of crop leaves according to the thermal imaging data, and judge whether the fumigation effect reaches the preset target; if the thermal imaging analysis calculation result shows that the preset target is not reached, dynamically adjust the fumigant spraying amount or spraying duration until the fumigation effect reaches the preset target, and then move to the next grid unit; if the thermal imaging analysis calculation result shows that the preset target is reached, the automated fumigation robot is controlled to move to the next grid unit to perform subsequent fumigation operation tasks; after the fumigation operation of all grid units is completed, the fumigation parameters and thermal imaging data of each grid unit are uploaded to the cloud data management platform, and data summary and analysis are performed to form a fumigation operation report for each prevention and control area.

[0039] Specifically, the automated fumigation robot control system obtains the geographical location information and area data of the control area through satellite positioning and mapping technology, and uses the image segmentation algorithm to grid the control area. According to the size and shape characteristics of the control area, the K-means clustering algorithm is used to divide each control area into several square or rectangular grid units of about 16 square meters. The system automatically matches the pest control knowledge base, and obtains the corresponding optimal fumigant type and dosage parameters according to the crop types of each control area, such as rice, wheat, corn, etc., and the degree of pest occurrence, such as mild, moderate, and severe. For example, the dosage of cypermethrin fumigants is 5 ml per square meter, and the parameters are written into the fumigation robot control system. According to the grid division scheme, the A* pathfinding algorithm is used to plan and generate the Z-shaped moving path of the fumigation robot in each grid unit. The moving path maintains a safe distance of 5 meters from the grid unit boundary. The length of the moving path in each grid unit is about 18 meters, which can effectively cover each grid unit area and avoid missing operations. The infrared thermal imager collects crop canopy thermal imaging data in real time, compares and analyzes the temperature gradient of adjacent crop leaves, and determines the area that may be missed if the temperature difference exceeds 5°C, and dynamically adjusts the robot's movement path. The fumigation robot is equipped with a high-precision GPS positioning system, which automatically navigates to the specified position of the grid unit. According to the crop distribution map and wind direction data in the grid unit, the fuzzy control algorithm is used to dynamically adjust the fumigant nozzle angle between 60°-120°, automatically identify the gap area between crop rows, and set it as the key fumigant spraying position. After arriving at the designated position, the fumigation robot uses the PID control algorithm to accurately control the electric valve opening and spraying time according to the optimal fumigant parameters. The fumigant spraying amount of each nozzle is controlled at 5-2 ml / m2, and the spraying time is controlled at 8-2 seconds. The infrared thermal imager collects crop canopy thermal imaging data every 5 seconds, transmits the data to the onboard processor, analyzes the data through the SVM support vector machine algorithm, calculates the average temperature drop rate of the crop leaves, and compares it with the preset standard model. If the temperature drop rate is lower than 5℃ / min, it is judged that the fumigation effect does not meet the standard, and the fumigant spraying amount is automatically increased by 5%, or the spraying time is extended by 2 seconds until the expected effect is achieved. If the temperature drop rate exceeds 5℃ / min, it is judged that the fumigation effect meets the standard, and the fumigation robot is controlled to directly enter the next grid unit to continue the operation. After the fumigation operation of all grid units is completed, the fumigation parameters and thermal imaging analysis data of each unit are uploaded to the cloud data management platform, and the data of each prevention and control area are summarized and analyzed using big data analysis technology to generate a fumigation operation effect evaluation report, and the machine learning algorithm is used to predict and warn the trend of pests and diseases, providing data support for the agricultural authorities to scientifically formulate green pest and disease control strategies.

[0040] Step S108, during the fumigation process, continuously collect audio data of pest chirping sounds in each control area, extract chirping sound intensity and frequency characteristic parameters, and evaluate the fumigation effect by comparing the chirping sound intensity and frequency characteristic parameters before and after fumigation; if the chirping sound parameters after fumigation do not reach the preset threshold, start the secondary fumigation program; if the preset threshold is reached, upload the chirping sound intensity and frequency characteristic parameters before and after fumigation to the cloud knowledge base.

[0041] According to the division of fumigation control areas, sound collection equipment is deployed in each area to collect audio data of pest chirping in real time; the audio data of pest chirping is preprocessed, including denoising and normalization, to obtain standardized audio data; the chirping intensity and frequency characteristic parameters are extracted from the standardized audio data to construct a characteristic vector of pest chirping before fumigation; the characteristic vector of pest chirping before fumigation is stored in a local database and uploaded to a cloud knowledge base for backup; during the fumigation process, the audio data of pest chirping in each control area is continuously collected, the chirping intensity and frequency characteristic parameters are extracted, and a characteristic vector of pest chirping during fumigation is constructed; the similarity between the characteristic vector of pest chirping during fumigation and the characteristic vector of pest chirping before fumigation is calculated to obtain a fumigation effect evaluation result. result; if the fumigation effect evaluation result does not reach the preset threshold, the repeated fumigation program is triggered, and the fumigation parameters are adjusted. During the repeated fumigation program, the audio data of the pest chirping sounds in each prevention and control area are continuously collected, the chirping sound intensity and frequency characteristic parameters are extracted, and the pest chirping sound feature vector in the repeated fumigation is constructed; the similarity between the pest chirping sound feature vector in the repeated fumigation and the pest chirping sound feature vector before fumigation is calculated to obtain the repeated fumigation effect evaluation result until the preset threshold is reached; if the repeated fumigation effect evaluation result reaches the preset threshold, the pest chirping sound feature vectors before and after fumigation are uploaded to the cloud knowledge base, and the pest chirping sound feature library is updated; the preset threshold is dynamically adjusted according to the historical fumigation data and the pest chirping sound feature library in the cloud knowledge base.

[0042] Specifically, according to the division of fumigation control areas, sound collection equipment is deployed in each area to collect audio data of pest chirping in real time. The sensitivity of the collection equipment can reach -40dB, the frequency response range is 20Hz-20kHz, the sampling rate is 41kHz, and the bit depth is 16bit. The collected audio data of pest chirping is preprocessed, including using an adaptive filtering algorithm to remove environmental noise, and normalizing the audio amplitude to map the audio amplitude to the [-1,1] interval to obtain standardized audio data. The chirping sound intensity and frequency characteristic parameters are extracted from the standardized audio data, and the audio signal is converted into a frequency domain representation using a short-time Fourier transform. The statistical features such as the mean and variance of the spectrum energy and frequency are extracted to construct the feature vector of the pest chirping sound before fumigation. The feature vector of the pest chirping sound before fumigation is stored in the local MySQL database and uploaded to the cloud MongoDB knowledge base for backup through the RESTfulAPI. During the fumigation process, the audio data of the pest chirping sound in each control area is continuously collected, the chirping sound intensity and frequency characteristic parameters are extracted, and the feature vector of the pest chirping sound during fumigation is constructed. The Euclidean distance is used to measure the similarity between the feature vector of pest chirping sounds during fumigation and the feature vector of pest chirping sounds before fumigation. If the similarity is lower than a preset threshold value, such as 8, the secondary fumigation procedure is triggered, and parameters such as fumigation agent concentration and duration are adjusted. Feature extraction and similarity calculation are re-executed until the preset threshold value is reached. If the fumigation effect evaluation result reaches the preset threshold, the feature vectors of pest chirping sounds before and after fumigation are uploaded to the cloud knowledge base, and the pest chirping sound feature library is updated. Based on the historical fumigation data and pest chirping sound feature library in the cloud knowledge base, the support vector machine algorithm is used to train and optimize the fumigation effect evaluation model. The hyperparameters such as the kernel function and penalty coefficient of the SVM are dynamically adjusted through grid search and cross-validation to improve the accuracy of the fumigation effect evaluation.

[0043] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the concept of the present application. For example, the above features are replaced with the technical features with similar functions disclosed in the present application (but not limited to) to form a technical solution.

Claims

1. A fumigation control method based on pest and disease trend prediction, It is characterized in that The method comprises: The audio data of insect chirping sounds in a preset frequency band is obtained through a sound sensor, the audio data of insect chirping sounds is processed by a Butterworth filter, and the audio data of insect chirping sounds is decomposed by a wavelet transform method to obtain audio data information of chirping sounds of different insect species; The fast Fourier transform algorithm is used to perform spectrum analysis on the audio data information to extract the frequency domain feature vector of each insect species' chirping sound. The frequency domain feature vector includes the fundamental frequency, harmonic energy distribution, chirping duration and chirping rhythm. Based on the extracted frequency domain feature vectors of the chirping sounds, the support vector machine algorithm is used to identify and classify the chirping sounds of the main local agricultural pests. By matching them with the preset insect species chirping sound feature template library, the pest species and life cycle information are determined. According to the information of pest species and life cycle, combined with the temperature and humidity data collected by the environmental temperature and humidity sensors, the logistic population growth model and Monte Carlo simulation algorithm are used to predict the growth curve of the pest population and the distribution data of insect population density. The distribution data of insect population density includes the mean value of insect population density and the confidence interval of insect population density. If the predicted mean value of insect population density exceeds the preset control threshold, a high-risk warning is output; The Kriging interpolation algorithm is used to spatially map the predicted insect population density distribution data, generate a spatial distribution map, determine the concentration and diffusion trend of pests in various areas of the field, and divide several control areas and buffer zones; For each control area, the particle swarm optimization algorithm is used to find the optimal fumigation plan between fumigation effect and cost, combining the area, pest species and insect population density distribution data of the control area. The optimal fumigation plan includes the combination of fumigant types, dosage ratio, fumigation time and temperature setting values. The optimal fumigation scheme parameters for each control area are input into the automated fumigation robot control system. By adjusting the fumigant spraying amount, spraying duration and moving path, fixed-point, quantitative and timed fumigation operations are carried out in each control area. The fumigation process is monitored by an infrared thermal imager, and the fumigation intensity is dynamically adjusted according to real-time feedback. During the fumigation process, audio data of pest chirping sounds in each prevention and control area is continuously collected, and the chirping sound intensity and frequency characteristic parameters are extracted. The fumigation effect is evaluated by comparing the chirping sound intensity and frequency characteristic parameters before and after fumigation. If the chirping sound parameters after fumigation do not reach the preset threshold, the secondary fumigation program is started. If the preset threshold is reached, the chirping sound intensity and frequency characteristic parameters before and after fumigation are uploaded to the cloud knowledge base.

2. The method according to claim 1, in, The method obtains the audio data of insect chirping sounds within a preset frequency band through a sound sensor, processes the audio data of insect chirping sounds using a Butterworth filter, and decomposes the audio data using a wavelet transform method to obtain audio data information of the chirping sounds of different insect species, including: A sound sensor is used to collect audio data of insect chirping sounds within a preset frequency band, and a sampling rate and sampling accuracy are set according to the frequency characteristics of the chirping sounds of different insect species to collect raw audio data, and the raw audio data is converted into a digital signal; Preprocessing the digital signal, identifying and marking the background noise in the original audio data through frequency domain analysis, and using a Butterworth filter to effectively suppress and remove the background noise according to the frequency distribution characteristics of the background noise; The processed digital signal is decomposed in the time-frequency domain by wavelet transform method, and the processed digital signal is decomposed into wavelet coefficients of different scales and frequencies; According to the energy distribution characteristics of the wavelet coefficients, the time-frequency domain characteristics of the chirping sounds of different insect species are determined to obtain the audio data information of the chirping sounds of different insect species.

3. The method according to claim 1, in, The fast Fourier transform algorithm is used to perform spectrum analysis on the audio data information to extract the frequency domain feature vector of the chirping sound of each insect species. The frequency domain feature vector includes the fundamental frequency, harmonic energy distribution, chirping duration and chirping rhythm, including: Get audio data information of the chirping sounds of different insect species; Performing spectrum analysis on the audio data using a fast Fourier transform algorithm to obtain a frequency domain feature vector, including fundamental frequency, harmonic energy distribution, call duration, and call rhythm; Determine the main frequency range of the chirping sound of each insect species according to the fundamental frequency in the frequency domain feature vector; Determine the frequency spectrum structure and energy concentration area of ​​the chirping sound through the harmonic energy distribution; According to the chirping duration, the average duration and variation range of the chirping sound of each insect species are obtained; Using the chirping rhythm, analyzing the time interval and regularity of the chirping sounds of insect species; If the dimension of the frequency domain feature vector is greater than a preset threshold, a principal component analysis method is used to perform dimensionality reduction processing; The frequency domain feature vectors are grouped by a clustering algorithm to obtain category labels of the chirping sounds of different insect species.

4. The method according to claim 1, in, The method uses a support vector machine algorithm to identify and classify the chirping sounds of local major agricultural pests based on the extracted chirping sound frequency domain feature vectors, and determines the type and life cycle information of the pests by matching with a preset insect species chirping sound feature template library, including: The extracted chirping sound frequency domain feature vectors are labeled, and a preset chirping sound feature template library of insect species is constructed by analyzing the labeled chirping sound frequency domain feature vectors; A support vector machine algorithm is used to train the frequency domain feature vector of the chirping sound, establish a chirping sound recognition and classification model, and optimize it; Inputting the frequency domain feature vector of the chirping sound collected in real time into the trained chirping sound recognition and classification model, and determining the type of the pest by matching it with the preset insect species chirping sound feature template library; According to the identified pest species and the preset pest life cycle characteristic information, the life cycle stage of the pest is analyzed and determined; If the identified pest species and the life cycle stage match the preset agricultural pest damage pattern, an early warning message is generated to indicate the agricultural pest damage risk; Send identification and classification results, pest life cycle information and early warning information to agricultural production managers; The preset insect species chirping sound feature template library and the chirping sound recognition and classification model are continuously optimized and updated.

5. The method according to claim 1, in, The method uses the logistic population growth model and Monte Carlo simulation algorithm to predict the growth curve of the pest population and the distribution data of the insect population density according to the information of the pest species and life cycle, combined with the temperature and humidity data collected by the environmental temperature and humidity sensor. The insect population density distribution data includes the mean value of the insect population density and the confidence interval of the insect population density. If the predicted mean value of the insect population density exceeds the preset control threshold, a high-risk warning is output, including: According to the pest species and life cycle information, a pest population growth model is established through a support vector machine algorithm, and key parameters of the pest population growth model are determined; Acquire temperature and humidity data collected by an environmental temperature and humidity sensor, perform preprocessing and feature extraction on the temperature and humidity data, and obtain quantitative indicators of environmental factors; Substituting the environmental factor quantitative index as input into the logistic population growth model to calculate the population growth rate and population size at the current moment; Using a Monte Carlo simulation algorithm, generating a random sample of population size according to the population size and the population growth rate; Performing statistical analysis on random samples of the population to obtain a probability distribution of the population, and calculating a mean insect population density and an insect population density confidence interval; Determine whether the mean insect population density exceeds a preset control threshold, and if the mean insect population density exceeds the preset control threshold, output high-risk warning information, including a warning level and recommended control measures; If the mean insect population density does not exceed the preset control threshold, the insect pest risk level is evaluated according to the insect population density confidence interval, and an insect monitoring plan is generated; Continuously monitor the changes in the quantitative indicators of the environmental factors and the population size, regularly update the population growth model and simulation prediction results, and optimize the preset control thresholds and the insect monitoring plan; Through visualization technology, insect population density distribution maps and population growth curves are generated to show the spatiotemporal dynamics of pest risks.

6. The method according to claim 1, in, The Kriging interpolation algorithm is used to perform spatial mapping on the predicted insect population density distribution data, generate a spatial distribution map, determine the pest aggregation degree and diffusion trend in each area of ​​the field, and divide a number of control areas and buffer zones, including: The Kriging interpolation algorithm is used to spatially map the predicted insect population density distribution data to obtain the spatial distribution map of insect population density distribution; According to the spatial distribution map of insect population density, determine the concentration and spread trend of pests in various areas of the field; According to the pest aggregation degree and the diffusion trend, a clustering algorithm is used to cluster the field area to divide it into several control areas and buffer zones; Obtain the geographical location information of each prevention and control area and buffer zone, and perform spatial positioning and boundary demarcation through geographic information system; Determine the control level and control measures for different areas based on the concentration and spread trend of pests in each control area and buffer zone; If the concentration of pests in the control area is greater than the preset concentration value and the spreading trend is greater than the preset value, chemical control measures are taken; If the concentration of pests in the control area is less than the preset concentration value and the diffusion trend is less than the preset value, physical control measures are taken; According to the geographical location and area of ​​the buffer zone, a spatial interpolation algorithm is used to calculate the insect population density distribution information in the buffer zone to determine whether prevention and control measures need to be taken; If the insect population density in the buffer zone exceeds a preset threshold, the buffer zone is divided into a control zone, and the control level and control measures are determined according to the insect population density; If the insect population density in the buffer zone does not exceed the preset threshold, the buffer zone will remain unchanged; Based on the divided prevention and control areas and buffer zones, a path planning algorithm is used to generate agricultural machinery operation paths and instruct agricultural machinery to apply pesticides.

7. The method according to claim 1, in, For each divided control area, the particle swarm optimization algorithm is used to solve the optimal fumigation scheme between fumigation effect and cost in combination with the area, pest species and insect population density distribution data of the control area. The optimal fumigation scheme includes the combination of fumigant types, dosage ratio, fumigation time and temperature setting values, including: Obtain the results of the division of control areas and the spatial distribution map of pest density in each area, calculate the area of ​​each control area, and obtain the geographical scope and area data of each control area; Obtain the pest species and insect population density distribution data of the divided control area, and build a multi-objective optimization model by using a particle swarm optimization algorithm in combination with the area data, wherein the objective function includes fumigation effect and cost; Determine candidate fumigant type combinations according to the pest types and the insect population density distribution data to form an initial particle group, each particle representing a fumigant type combination and a corresponding dosage ratio, fumigation time and temperature setting value; Calculating the fitness value of each particle, the fitness value combines the two objective functions of fumigation effect and cost, and is weighted according to the area of ​​the control area; Update the position and speed of the particles, and obtain a new particle swarm based on the fitness value of the particles using the update formula of the particle swarm optimization algorithm; Determine whether the new particle swarm meets the termination condition, and if so, output the global optimal solution as the optimal fumigation solution; If the termination condition is not met, the position and speed of the particle are updated, and a new particle swarm is obtained according to the fitness value of the particle using the update formula of the particle swarm optimization algorithm; According to the optimal fumigation plan, determine the type combination, dosage ratio, fumigation time and temperature setting value of each control area; Acquire environmental temperature and humidity data, and dynamically adjust the fumigation time and the temperature setting value; By monitoring the change information of pest species and insect population density, the fumigation effect is evaluated, and the evaluation results are fed back to the multi-objective optimization model to dynamically update the parameters of the multi-objective optimization model; Based on the evaluation results of the fumigation effect and the cost accounting data, a multi-objective optimization model is used to continuously optimize the fumigation plan and manage the fumigation process.

8. The method according to claim 1, in, The optimal fumigation scheme parameters of each control area are input into the automatic fumigation robot control system, and the fixed-point, quantitative and timed fumigation operation is carried out on each control area by adjusting the fumigant spraying amount, spraying time and moving path, and the fumigation process is monitored by an infrared thermal imager, and the fumigation intensity is dynamically adjusted according to real-time feedback, including: Obtain the geographical location information and area data of each control area, generate a grid division scheme for each control area according to the area size and shape characteristics of the control area, and divide each control area into a number of grid units; According to the degree of occurrence of pests and diseases and crop types in each control area, the pest and disease control knowledge base is matched to obtain the optimal fumigation scheme parameters for each control area, and the optimal fumigation scheme parameter data is input into the automatic fumigation robot control system; According to the grid division scheme, a moving path of the automated fumigation robot in each grid unit is generated, and the moving path needs to fully cover the area of ​​each grid unit; Controlling the automated fumigation robot to navigate and move to a designated position of each grid unit according to the moving path, and determining a spraying position of the fumigant according to a crop distribution map within the grid unit; After the automated fumigation robot reaches the spraying position, it adjusts the fumigant spraying amount and spraying duration according to the fumigant type and dosage parameters of the optimal fumigation scheme, and performs fumigation operations on the crops in the grid unit; The infrared thermal imager is used to collect thermal imaging data of crop canopies in real time, and the temperature change trend of crop leaves is analyzed and calculated based on the thermal imaging data to determine whether the fumigation effect has reached the preset target; If the thermal imaging analysis calculation result shows that the preset target has not been achieved, the fumigant spraying amount or spraying duration is dynamically adjusted until the fumigation effect reaches the preset target, and then moves to the next grid unit; If the thermal imaging analysis calculation result shows that the preset target is achieved, the automated fumigation robot is controlled to move to the next grid unit to perform subsequent fumigation operation tasks; After the fumigation operation of all grid units is completed, the fumigation parameters and thermal imaging data of each grid unit will be uploaded to the cloud data management platform for data summary and analysis to form a fumigation operation report for each prevention and control area.

9. The method according to claim 1, in, During the fumigation process, the audio data of the pest chirping sound in each control area is continuously collected, the chirping sound intensity and frequency characteristic parameters are extracted, and the fumigation effect is evaluated by comparing the chirping sound intensity and frequency characteristic parameters before and after fumigation. If the chirping sound parameters after fumigation do not reach the preset threshold, the secondary fumigation program is started. If the preset threshold is reached, the chirping sound intensity and frequency characteristic parameters before and after fumigation are uploaded to the cloud knowledge base, including: According to the division of fumigation control areas, sound collection equipment is deployed in each area to collect audio data of pest chirping sounds in real time; Preprocessing the audio data of the pest chirping sound, including denoising and normalization, to obtain standardized audio data; Extracting chirping sound intensity and frequency characteristic parameters from the standardized audio data to construct a chirping sound feature vector of pests before fumigation; The feature vector of the pest chirping sound before fumigation is stored in a local database and uploaded to a cloud knowledge base for backup; During the fumigation process, the audio data of pest chirping sounds in each control area is continuously collected, the characteristic parameters of chirping sound intensity and frequency are extracted, and the characteristic vector of pest chirping sounds during fumigation is constructed; Calculate the similarity between the characteristic vector of the pest chirping sound during fumigation and the characteristic vector of the pest chirping sound before fumigation to obtain a fumigation effect evaluation result; If the fumigation effect evaluation result does not reach the preset threshold, the repeated fumigation procedure is triggered, and the fumigation parameters are adjusted. During the repeated fumigation procedure, the audio data of the pest chirping sounds in each control area are continuously collected, the chirping sound intensity and frequency characteristic parameters are extracted, and the characteristic vector of the pest chirping sounds in the repeated fumigation is constructed; Calculate the similarity between the characteristic vector of the pest chirping sound during the repeated fumigation and the characteristic vector of the pest chirping sound before the fumigation to obtain a repeated fumigation effect evaluation result until the preset threshold is reached; If the repeated fumigation effect evaluation result reaches the preset threshold, the pest chirping sound feature vectors before and after fumigation are uploaded to the cloud knowledge base to update the pest chirping sound feature library; The preset threshold is dynamically adjusted according to the historical fumigation data and the pest chirping sound feature library in the cloud knowledge base.

Images