Pesticide spraying detection method based on machine learning

Abstract

The invention discloses a pesticide spraying detection method based on machine learning, and relates to the technical field of pesticide spraying detection. Sprinkling irrigation measures are finely customized according to pesticide influence grades, water is reasonably increased, an irrigation mode is optimized and fertilizer is supplemented timely when the sprinkling irrigation measures are deficient, so that soil moisture and nutrients are ensured to adapt to crop requirements, and healthy and strong growth of crops is promoted; when the amount is excessive, sprinkling irrigation is scientifically reduced, a nozzle is adjusted, and a relieving agent is added, so that the excessive effect of the pesticide in the soil is relieved, a suitable growth environment is created for crops, and the utilization rate of water resources and fertilizer is increased; the use amount, the application mode and the application time of various fertilizers are accurately adjusted. Nutrients are supplemented when the fertilizer is deficient, and the fertilizer is reasonably controlled when the fertilizer is excessive, so that the nutrient requirements of crop growth are met, the problem that the crop is aggravated due to poor interaction of the fertilizer and pesticide is avoided, the goal of increasing production and income is finally achieved, and the modern agriculture is promoted to be intelligentized and refined.

Description

Technical Field

[0001] The present invention relates to the technical field of pesticide spraying detection, and specifically relates to a pesticide spraying detection method based on machine learning. Background Art

[0002] Modern agriculture has an urgent demand for intelligent and refined management. With the progress of technology, the agricultural field is eager to introduce high-tech to improve production efficiency and benefits. The traditional management mode of manual field inspection and experience-based judgment can no longer meet the needs of large-scale and modern farms for real-time and accurate decision-making. There is an urgent need to utilize the powerful data processing and model construction capabilities of machine learning to deeply explore the internal laws of the interaction between crops and pesticides, so as to realize the transformation of pesticide spraying from "experience-driven" to "data-driven", thereby ensuring the sustainable development of agriculture, helping farmers increase production and income, and enhancing the overall competitiveness of the agricultural industry.

[0003] The prior art, such as a pesticide spraying control method, system, device and storage medium disclosed in the invention patent application with the publication number of CN117814202A, the method includes: obtaining target crop information; according to the target crop information, controlling a target sprayer to determine a corresponding spraying mode and obtaining change information of the target crop; according to the change information, judging whether the dosage of the pesticide for the target crop reaches a dosage warning threshold corresponding to the target crop; if the dosage reaches the dosage warning threshold, controlling the target sprayer to close the spraying mode. This application has the beneficial effect of scientifically using pesticides, thereby avoiding environmental pollution and ensuring the edible safety of agricultural products.

[0004] In view of the above solution, the inventors of the present application found that the above technology has at least the following technical problems: 1. Most of the prior art relies on experience or simple pesticide instruction manuals to estimate the active period of pesticides, lacking systematic comprehensive consideration. It only recommends based on the conventional action duration of pesticides, without fully considering the unique metabolic characteristics of different crop varieties at different growth stages, the differences in the absorption and transformation of pesticides, and the precise impact of the actual spraying amount. This is likely to lead to inaccurate judgment of the active period of pesticides, resulting in premature supplementary spraying of pesticides, increasing costs and the risk of pesticide residues, or spraying pesticides too late, causing pests and diseases to get out of control and endangering crop growth. Due to the lack of a powerful database to support the comparison of multi-factor information, it is difficult to cope with complex and changeable actual planting scenarios. For example, when encountering newly introduced crop varieties, special climate conditions affecting the growth rhythm of crops, or using new pesticide formulations, the prior art cannot quickly and accurately adapt, and cannot give a targeted pesticide active period, making pest control blind and affecting the stability of agricultural output.

[0005] 2. The existing technical evaluation methods tend to be single-minded and often focus on a single indicator. For example, they only focus on the pest and disease control rate to measure the effect of pesticides, ignoring the comprehensive situation of the growth of crops themselves, physiological and biochemical status, and the impact of pesticides on the quality of agricultural products. This may lead to misjudgment. It seems that pests and diseases are under control, but the crop growth is slow and the fruit quality is reduced, which ultimately affects the economic benefits and reduces the market competitiveness of agricultural products. Lack of dynamic, continuous monitoring and quantitative analysis capabilities. Most of the existing technologies are staged static observations. It is impossible to obtain crop growth assessment values, physiological and biochemical assessment values, and quality assessment values ​​at multiple collection time points like machine learning methods, and comprehensively quantify the interaction between pesticides and crops through scientific model formulas. It is difficult to track the entire process of pesticide action in real time, it is difficult to capture the subtle crop reactions caused by changes in pesticide efficacy in a timely manner, and it is impossible to provide real-time and effective basis for subsequent precision farming operations.

[0006] 3. Existing sprinkler irrigation and fertilization measures lack close coordination and linkage with the impact of pesticides. They are usually operated according to a fixed irrigation and fertilization schedule without considering the real-time performance status of pesticides. For example, when the pesticide is excessive, large-scale irrigation and fertilization are still carried out as usual, which increases the accumulation of pesticide residues in the soil and the burden on crops; when pesticides are insufficient, irrigation and fertilization are not adjusted in time to promote crop resistance to diseases and pests, resulting in insufficient motivation for crop growth, making it difficult to achieve accurate resource allocation, and reducing agricultural production efficiency. Traditional methods are difficult to flexibly adjust agricultural operation details according to the real-time needs of crops. It is impossible to finely adjust the amount of sprinkler water, frequency, and nozzle angle according to the level of pesticide impact like machine learning methods, and accurately control the type, amount, application method and timing of fertilizers, resulting in waste of water resources and fertilizers, and it is impossible to meet the specific needs of crops at different stages for growth environment and nutrients, hindering the development of modern agriculture towards efficiency and intelligence. Summary of the invention

[0007] In view of the above-mentioned technical deficiencies, the object of the present invention is to provide a pesticide spraying detection method based on machine learning.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solutions: The present invention provides a pesticide spraying detection method based on machine learning, including: Step 1, analysis of the pesticide active period: After the pesticide spraying of the target crop is completed, the pesticide efficacy influencing factors corresponding to the target crop are obtained, and then the pesticide active period corresponding to the target crop is analyzed, and when the pesticide active period of the target crop is reached, a number of collection points are set, and then the pesticide and crop interaction evaluation coefficient corresponding to the target crop is analyzed.

[0009] Step 2: Evaluation of the impact of pesticides on crops: Based on the pesticide-crop interaction evaluation coefficient corresponding to the target crop, the efficacy status of the pesticide corresponding to the target crop on the crop is evaluated. The efficacy status includes good, excess and deficiency.

[0010] Step 3. Coordination of sprinkler irrigation and fertilization measures: If the efficacy state of the pesticide corresponding to the target crop on the crop is excessive or insufficient, an excessive warning prompt and an insufficient warning prompt are given, so as to analyze the level of the impact of the pesticide on the target crop, and then analyze the sprinkler irrigation measures and fertilization measures corresponding to the target crop.

[0011] Preferably, the analysis of the pesticide active period corresponding to the target crop is as follows: Obtain the pesticide efficacy influencing factors corresponding to the target crop. The pesticide efficacy influencing factors include crop variety, crop growth stage, pesticide type, and pesticide spraying amount. Compare the crop variety, crop growth stage, pesticide type, and pesticide spraying amount corresponding to the target crop with those corresponding to each pesticide active period in the database. If the crop variety, crop growth stage, pesticide type, and pesticide spraying amount corresponding to the target crop are the same as those corresponding to a certain pesticide active period in the database, then take the pesticide active period in the database as the pesticide active period corresponding to the target crop.

[0012] Preferably, the evaluation of the efficacy state of the pesticide corresponding to the target crop on the crop is as follows: B1. Compare the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop with the set standard evaluation coefficient range of the interaction between the pesticide and the crop of the standard target crop.

[0013] B2. If the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop is within the set standard evaluation coefficient range of the interaction between the pesticide and the crop of the standard target crop, it indicates that the efficacy state of the pesticide corresponding to the target crop on the crop is good.

[0014] B3. If the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop is greater than the maximum value within the set standard evaluation coefficient range of the interaction between the pesticide and the crop of the standard target crop, it indicates that the efficacy state of the pesticide corresponding to the target crop on the crop is excessive.

[0015] B4. If the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop is less than the minimum value within the set standard evaluation coefficient range of the interaction between the pesticide and the crop of the standard target crop, it indicates that the efficacy state of the pesticide corresponding to the target crop on the crop is insufficient.

[0016] Preferably, the level of the impact of the pesticide on the target crop is analyzed as follows: C1. If the efficacy state of the pesticide corresponding to the target crop on the crop is insufficient, the difference is calculated between the interaction evaluation coefficient of the pesticide corresponding to the target crop and the crop and the minimum value within the set standard interaction evaluation coefficient range of the pesticide corresponding to the standard target crop and the crop, obtaining the difference between the interaction evaluation coefficient of the pesticide corresponding to the target crop and the crop and the minimum value within the set standard interaction evaluation coefficient range of the pesticide corresponding to the standard target crop and the crop. If this difference is greater than or equal to the deficiency threshold, it is recorded as the first level of deficiency; if this difference is less than the deficiency threshold, it is recorded as the second level of deficiency.

[0017] C2. If the efficacy state of the pesticide corresponding to the target crop on the crop is excessive, the difference is calculated between the interaction evaluation coefficient of the pesticide corresponding to the target crop and the crop and the maximum value within the set standard interaction evaluation coefficient range of the pesticide corresponding to the standard target crop and the crop, obtaining the difference between the interaction evaluation coefficient of the pesticide corresponding to the target crop and the crop and the maximum value within the set standard interaction evaluation coefficient range of the pesticide corresponding to the standard target crop and the crop. If this difference is less than or equal to the excess threshold, it is recorded as the first level of excess; if this difference is greater than the excess threshold, it is recorded as the second level of excess.

[0018] Preferably, the sprinkler irrigation measures corresponding to the target crop are analyzed as follows: D1. If the level of the impact of the pesticide on the target crop is the first level of deficiency, increase the sprinkler irrigation water volume, which is 10%-15% higher than the conventional sprinkler irrigation volume, ensure that the soil humidity is maintained at 70%-75% of the field capacity. At the same time, adopt a sprinkler irrigation mode with small water droplets and high frequency, sprinkle once every 2-3 hours, and each sprinkler irrigation lasts for 10-15 minutes. Adjust the sprinkler irrigation time and choose to carry out sprinkler irrigation in the early morning or evening.

[0019] D2. If the level of the impact of the pesticide on the target crop is the second level of deficiency, significantly increase the sprinkler irrigation water volume, which is 20%-30% higher than the normal sprinkler irrigation volume, ensure that the soil humidity quickly reaches 75%-80% of the field capacity. Adopt a combination of flood irrigation and sprinkler irrigation. First, carry out a short-time flood irrigation every 30-45 minutes, and then switch to sprinkler irrigation to maintain the surface soil humidity, sprinkle once every 1 hour. At the same time, add amino acid water-soluble fertilizer with a concentration controlled at 0.2%-0.3%.

[0020] D3. If the level of the impact of the pesticide on the target crop is the first level of excess, immediately reduce the sprinkler irrigation water volume, reduce the sprinkler irrigation frequency by 30%-40%, and shorten the single sprinkler irrigation time by 20%-30%, so that the soil humidity gradually drops to 60%-65% of the field capacity. Adjust the angle of the sprinkler irrigation nozzle to be higher than the crop canopy and spray in a mist form.

[0021] D4. If the level of the impact of the pesticide on the target crop is rated as over-dose level 2, suspend sprinkler irrigation for 1 - 2 days to reduce the soil humidity to 50% - 55% of the field capacity. Then, adopt a sprinkler irrigation strategy of small amounts but multiple times. The amount of each sprinkler irrigation is only 20% - 30% of the normal amount, and the interval is 4 - 6 hours. Add a composite solution containing microbial inoculants to the sprinkler irrigation water, with a concentration of 0.1% - 0.2%.

[0022] Preferably, analyze the fertilization measures corresponding to the target crop. The specific analysis process is as follows: E1. If the level of the impact of the pesticide on the target crop is rated as deficiency level 1, increase the application rate of nitrogen fertilizer, increasing it by 10% - 15% on the basis of the original planned nitrogen fertilizer application rate. At the same time, apply trace element fertilizers in combination, with a concentration of 0.1% - 0.2%, using foliar spraying. Adjust the fertilization time and frequency, advance the originally planned topdressing time by 3 - 5 days. On the basis of originally fertilizing once every two weeks, shorten the interval to once every 10 - 12 days. Adopt a combination of basal fertilizer and topdressing. The basal fertilizer is mainly organic fertilizer, and the application rate per mu is increased by 5% - 10%. The topdressing selects quick-acting chemical fertilizers and is applied by drip irrigation or shallow burial.

[0023] E2. If the level of the impact of the pesticide on the target crop is rated as deficiency level 2, significantly increase the fertilizer application rate. The nitrogen fertilizer application rate is increased by 20% - 30% on the original basis. At the same time, apply biological bacterial fertilizer, with a dosage of 10 - 15 kg per mu. In addition, add fertilizers containing amino acid and humic acid components, with a concentration of 0.3% - 0.5% through foliar spraying. Adjust the fertilization time and frequency, shorten the original fertilization interval to 7 - 10 days, and conduct a centralized fertilization in advance before the high-incidence period of plant diseases and insect pests. The fertilization time is selected on a cloudy day or in the evening. Adopt a three-dimensional fertilization method combining foliar fertilization and root fertilization. For root fertilization, use the deep application method to apply the fertilizer to the dense root layer. For foliar fertilization, conduct it once every 5 - 7 days.

[0024] E3. If the level of the impact of the pesticide on the target crop is rated as over-dose level 1, reduce the application of nitrogen fertilizer, reducing the original nitrogen fertilizer application rate by 10% - 15%. Increase the application rate of potassium fertilizer, increasing it by 10% - 15% on the basis of the original planned potassium fertilizer application rate. At the same time, add fertilizers with the function of regulating crop growth. Adjust the fertilization time and frequency, postpone the originally planned topdressing time by 3 - 5 days, and reduce the fertilization frequency. On the basis of originally fertilizing once every two weeks, extend the interval to once every 16 - 18 days. Adopt the shallow application and scattered application methods to evenly spread the fertilizer in the shallow soil around the crop roots. For foliar fertilizer, reduce the concentration and application frequency, reduce the foliar fertilizer concentration to 0.05% - 0.1%, and adjust the application frequency to once every 10 - 15 days.

[0025] E4. If the level of the impact of the pesticide on the target crop being analyzed is excessive level 2, significantly reduce the amount of fertilizer used. Reduce the amount of nitrogen fertilizer by 30% - 40% based on the original amount. At the same time, apply fertilizers with detoxification and repair functions, adjust the fertilization time and frequency, suspend fertilization for 1 - 2 weeks, reduce the fertilization frequency to once a month, and adopt the method of local fertilization by opening shallow trenches for fertilization at a distance of 10 - 15 cm around the roots.

[0026] The beneficial effects of the present invention are as follows: 1. In the embodiments of the present invention, through the sprinkler irrigation measures being finely customized according to the pesticide impact level, when it is insufficient, reasonably increase the water volume, optimize the irrigation mode and timely supply fertilizer to ensure that the soil moisture and nutrients meet the crop requirements and promote the healthy growth of the crops; when it is excessive, scientifically reduce sprinkler irrigation, adjust the nozzles and add mitigation agents to reduce the excessive effect of pesticides in the soil, create a suitable growth environment for the crops, improve the utilization rate of water resources and fertilizers, and enhance the agricultural production efficiency. At the same time, the fertilization plan cooperates synergistically with the pesticide impact. According to the deficiency or excess conditions of different levels, accurately adjust the dosage, application method and timing of various fertilizers such as nitrogen fertilizer and potassium fertilizer. Supply nutrients when lacking fertilizer and reasonably control fertilizer when excessive, which not only meets the nutritional needs of crop growth but also avoids the adverse interaction between fertilizer and pesticide from exacerbating crop problems, realizes precise nutrient supply, helps the crops grow healthily, and finally achieves the goal of increasing production and income, promoting the modern agriculture to move towards intelligence and refinement.

[0027] 2. In the embodiments of the present invention, by comprehensively comparing multiple factors such as crop variety, growth stage, pesticide type and spraying amount with the database information, the effective action duration of the pesticide on the target crop can be accurately locked. This avoids the problems of repeated pesticide application or premature or late pesticide application caused by misjudgment of the pesticide active period, reduces pesticide waste, lowers the agricultural production cost, and at the same time ensures the continuous effectiveness of pesticide protection during the critical period of high incidence of pests and diseases, guaranteeing the growth safety of the crops. Build a comprehensive and refined evaluation system for the interaction between pesticides and crops, considering crop growth, physiological and biochemical and quality indicators from multiple dimensions. No longer limited to the traditional single - index judgment, this comprehensive evaluation accurately reflects the actual efficacy of pesticides, provides a solid data basis for the subsequent adjustment of farming operations, and makes agricultural production decisions more scientific and targeted.

[0028] 3. In the embodiments of the present invention, through a hierarchical over-dose warning mechanism, from the prominent reminder of the APP in the first-level warning, the red marking of the plot and the display of pesticide damage symptoms, to the on-site broadcast of the smart speaker, and then to the strong reminder of the APP, function locking and emergency warning of the drone in the second-level warning, it ensures that farmers can detect the pesticide over-dose crisis in the first time in all-round and multi-angle ways. Quickly respond and adjust farming operations such as irrigation and fertilization, minimize the loss of pesticide damage, prevent irreversible damage to crops, and protect the yield and quality of crops. The lack of warning also operates finely in grades. Through means such as APP push, plot color marking, field prompts on the electronic display screen and voice call urging, farmers are timely informed of the insufficient effect of pesticides. It prompts farmers to optimize irrigation and fertilization strategies in time, ensures that the growth of crops is not hindered due to poor pesticide effects, and maintains a good growth trend. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0030] Figure 1 It is a flowchart of the implementation steps of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0032] As shown in the embodiments of the present invention Figure 1 A pesticide spraying detection method based on machine learning includes: Step 1. Analysis of the pesticide active period: After the pesticide spraying of the target crop is completed, obtain the pesticide efficacy influencing factors corresponding to the target crop, and then analyze the pesticide active period corresponding to the target crop. When the pesticide active period of the target crop arrives, set several collection points, and then analyze the interaction evaluation coefficient between the pesticide corresponding to the target crop and the crop.

[0033] It should be noted that the pesticide active period refers to the time period during which the pesticide can effectively play its expected functions such as preventing and controlling pests and diseases and regulating crop growth after being sprayed on the target crop.

[0034] In a specific embodiment, the analysis of the pesticide active period corresponding to the target crop is carried out as follows: Obtain the factors affecting the pesticide efficacy corresponding to the target crop. The factors affecting the pesticide efficacy include crop variety, crop growth stage, pesticide type, and pesticide spraying amount. Compare the crop variety, crop growth stage, pesticide type, and pesticide spraying amount corresponding to the target crop with those corresponding to each pesticide active period in the database. If the crop variety, crop growth stage, pesticide type, and pesticide spraying amount corresponding to the target crop are the same as those corresponding to a certain pesticide active period in the database, then take the pesticide active period in the database as the pesticide active period corresponding to the target crop.

[0035] It should be noted that high-definition cameras are installed in the field to take crop images regularly. Deep learning algorithms are used to analyze the images to identify the morphological characteristics of the crops, and they are compared with the pre-established crop variety image database to automatically determine the crop variety. Various sensors are deployed in the field, such as light sensors, temperature sensors, humidity sensors, and plant conductivity sensors, etc. Crops at different growth stages have different requirements for environmental conditions and physiological responses. By analyzing the data collected by the sensors in real time and combining with the crop growth model, the growth stage of the crops can be inferred. When purchasing pesticides, farmers or agricultural production units are required to record in detail information such as the name, dosage form, active ingredient, manufacturer, purchase date, and purchase quantity of the pesticides. These records can be stored in paper files or electronic databases for easy query and traceability. Most modern precision agriculture pesticide application equipment is equipped with flow sensors, pressure sensors, and intelligent control systems. During the pesticide spraying process, these devices automatically record information such as the spraying flow rate, spraying time, and nozzle working status of the pesticides.

[0036] It should also be noted that machine learning is applied in the pesticide active period analysis process. The determination of the pesticide active period of the target crop is achieved by comparing the crop variety, growth stage, pesticide type, and spraying amount of the target crop with the database data. The regression algorithm of machine learning is used to optimize this process. A large amount of crop sample data is collected. In addition to crop variety, growth stage, pesticide type, and spraying amount, environmental factors are also included as features, and the pesticide active period is used as the target variable. These data are cleaned, preprocessed, and feature selected to improve the training effect of the model. Algorithms such as linear regression, decision tree regression, or random forest regression are used to train the data. By continuously adjusting the model parameters, the model can learn the complex relationship between each feature and the pesticide active period. When there is a new target crop, its relevant features are input into the trained model, and the model can predict the pesticide active period corresponding to this crop, which can obtain more accurate and reliable results compared with simple comparison.

[0037] In another specific embodiment, the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop is analyzed as follows: Obtain the crop growth evaluation value, crop physiological and biochemical evaluation value, and crop quality evaluation value corresponding to the target crop at each collection time point, and denote the crop growth evaluation value, crop physiological and biochemical evaluation value, and crop quality evaluation value corresponding to the target crop at each collection time point as , and respectively, where k represents the number corresponding to each collection time point, , q is an arbitrary integer greater than 2 and represents the set of all collection time points, and substitute the crop growth evaluation value, crop physiological and biochemical evaluation value, and crop quality evaluation value corresponding to the target crop at each collection time point into the calculation formula: to obtain the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop, where , and are the standard crop growth evaluation value, standard crop physiological and biochemical evaluation value, and standard crop quality evaluation value corresponding to the set target crop respectively, , , are the weight factors corresponding to the crop growth evaluation value, crop physiological and biochemical evaluation value, and crop quality evaluation value of the set target crop respectively.

[0038] It should be noted that , , are all greater than 0 and less than 1.

[0039] It should also be noted that senior experts in the agricultural field are consulted, and based on their profound understanding of the growth characteristics of different crops accumulated in long-term practice, the standard values are calibrated and improved. At the same time, a large number of authoritative agricultural literature and scientific research reports are consulted. These materials often contain the growth parameters of specific crops under ideal experimental conditions and can be used as supplementary references. Combining these data with the local actual situation, agricultural experts determine the standard crop growth evaluation value, standard crop physiological and biochemical evaluation value, and standard crop quality evaluation value corresponding to the suitable target crop.

[0040] Once again, it is necessary to collect a large amount of sample data including crop growth, physiological and biochemical, and quality indicators, and use the PCA method to reduce the dimension of the data. By analyzing the variance contribution rate of each principal component, it is determined which indicators have a strong ability to explain data variation, that is, they have a great impact on the evaluation of pesticide-crop interactions. PCA mines the importance of indicators based on the inherent structure of the data itself, avoids the subjectivity of human judgment, and provides an objective basis for weight setting. It is especially suitable for multivariate complex evaluation systems to set the weight factors corresponding to the target crop growth evaluation value, the weight factors corresponding to the crop physiological and biochemical evaluation value, and the weight factors corresponding to the crop quality evaluation value.

[0041] In another specific embodiment, the crop growth assessment value, crop physiological and biochemical assessment value and crop quality assessment value corresponding to the target crop at each collection time point are obtained, and the specific acquisition process is as follows: A1. Obtain the crop growth indicators, crop physiological and biochemical indicators and crop quality indicators corresponding to the target crop at each collection time point, the crop growth indicators include biomass volume, branch number change rate and chlorophyll content, the crop physiological and biochemical indicators include photosynthetic pigment content, carboxylation efficiency and antioxidant enzyme activity change rate, and the crop quality indicators include pesticide residues, nutrient content change rate and shape regularity.

[0042] It should be noted that drones equipped with multispectral cameras fly over farmland at regular intervals according to preset routes. Multispectral cameras can capture the reflectance characteristics of crops in different spectral bands, and build biomass estimation models by analyzing these characteristics. High-definition cameras are installed at key locations in the field. The cameras are equipped with intelligent image recognition software, and use deep learning algorithms to analyze crop images in real time. Starting from the seedling stage of crops, the system automatically takes photos of crops at predetermined time intervals, identifies and marks the main stem and branches of each crop, and accurately counts the changes in the number of branches by comparing image data at different time points. Chlorophyll sensors are distributed and installed in the field. The sensors use advanced fluorescence detection or light absorption principles to measure the relative chlorophyll content of crop leaves in real time.

[0043] It should also be noted that a portable near-infrared spectrometer is used to directly scan and detect crop leaves in the field. Near-infrared spectroscopy can reflect the chemical bond vibration information of photosynthetic pigments. By establishing a mathematical model between the spectrum and the content of photosynthetic pigments, the content of photosynthetic pigments can be quickly estimated. Multiple intelligent photosynthetic gas exchange monitoring units are set up in the field, and each unit is equipped with a high-precision portable photosynthesizer and an automated control system. When carboxylation efficiency data needs to be collected, the control system automatically adjusts the position of the leaf chamber of the photosynthesizer to accurately fix it on the pre-selected crop leaf to ensure good sealing. Subsequently, the gradient changes of environmental factors such as light intensity and carbon dioxide concentration are automatically set according to the preset program, and the net photosynthetic rate of the leaf is measured in real time under different conditions. Through the built-in data processing module, based on the relationship curve between the net photosynthetic rate and the carbon dioxide concentration, the carboxylation efficiency is automatically calculated, and the result is uploaded to the cloud server. Using microfluidic chip technology, steps such as enzyme solution extraction, reaction, and detection are integrated on a tiny chip. In the field, only a small amount of leaf samples need to be collected, processed and then injected into the microfluidic chip. The microchannels and micro reaction chambers inside the chip automatically complete the mixing and reaction of enzymes and reagents according to the preset program, and the absorbance change is detected by the micro sensors integrated on the chip to quickly obtain the antioxidant enzyme activity result.

[0044] Once again, it should be noted that an in-situ rapid pesticide residue detection system based on technologies such as immunoassay and biosensors is directly deployed in the fields. When agricultural products are approaching the harvest period, according to the preset sampling rules, the system automatically collects samples of fruits, leaves or other edible parts. After simple processing of the samples, they are sent to the detection module. The detection module uses the principle of specific binding of antigen-antibody or enzyme-catalyzed reactions, etc., to quickly detect the pesticide residue content in the samples. Combining an intelligent image acquisition device and a near-infrared spectrometer to construct a fusion monitoring system. The image acquisition device obtains image information such as the appearance and color of agricultural products, and the near-infrared spectrometer collects spectral information of internal nutritional components. Using 3D structured light scanning technology, after the harvest or in the late growth stage of agricultural products, they are quickly scanned to obtain a 3D model of the agricultural products. Through the analysis of the 3D model, the shape parameters of agricultural products can be measured more accurately.

[0045] A2. Input the biomass accumulation, branch number change rate, and chlorophyll content of the target crop corresponding to each collection time point into the crop growth evaluation value analysis model, and output the crop growth evaluation value of the target crop corresponding to each collection time point.

[0046] It should be noted that the analysis process of the crop growth evaluation value of the target crop corresponding to each collection time point is as follows: The biomass accumulation, branch number change rate, and chlorophyll content of the target crop corresponding to each collection time point are normalized, and the biomass accumulation, branch number change rate, and chlorophyll content of the target crop corresponding to each collection time point after processing are respectively denoted as , and , substitute into the analysis formula to obtain the crop growth evaluation values corresponding to the target crop at each collection time point .

[0047] A3. Input the photosynthetic pigment content, carboxylation efficiency, and antioxidant enzyme activity change rates corresponding to the target crop at each collection time point into the crop physiological and biochemical evaluation value analysis model, and output the crop physiological and biochemical evaluation values corresponding to the target crop at each collection time point.

[0048] It should be noted that the crop physiological and biochemical evaluation values corresponding to the target crop at each collection time point are obtained by analyzing according to the above analysis process of the crop growth evaluation values corresponding to the target crop at each collection time point.

[0049] A4. Input the pesticide residue amount, nutrient content change rate, and shape regularity corresponding to the target crop at each collection time point into the crop quality evaluation value analysis model, and output the crop quality evaluation values corresponding to the target crop at each collection time point.

[0050] It should be noted that the crop quality evaluation values corresponding to the target crop at each collection time point are obtained by analyzing according to the above analysis process of the crop growth evaluation values corresponding to the target crop at each collection time point.

[0051] It should be noted that in the link of obtaining crop evaluation values, machine learning is applied, and the crop growth, physiological and biochemical, and quality evaluation value analysis models are used to obtain evaluation values. These models are constructed using algorithms. A large amount of data on crop growth indicators, crop physiological and biochemical indicators, and crop quality indicators of different crops at each collection time point are collected, and each sample is labeled with the corresponding evaluation value. A model is constructed, and the labeled data is used to train the model. The weights and biases are continuously adjusted through the backpropagation algorithm to make the model output as close as possible to the labeled evaluation value. In practical applications, the various indicators of the collected target crop are input into the trained model, and the model can output the corresponding evaluation value, providing more accurate data for subsequent analysis.

[0052] Step 2. Evaluation of the impact of pesticides on crops: According to the pesticide-crop interaction evaluation coefficient corresponding to the target crop, further evaluate the efficacy state of the pesticide corresponding to the target crop on the crop. The efficacy state includes good, excessive, and insufficient.

[0053] In a specific embodiment, the specific analysis process for evaluating the efficacy state of the pesticide corresponding to the target crop on the crop is as follows: B1. Compare the pesticide-crop interaction evaluation coefficient corresponding to the target crop with the set standard pesticide-crop interaction evaluation coefficient interval for the target crop.

[0054] B2. If the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop is within the set standard evaluation coefficient range of the interaction between the pesticide and the crop corresponding to the target crop, it indicates that the efficacy state of the impact of the pesticide corresponding to the target crop on the crop is good.

[0055] B3. If the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop is greater than the maximum value within the set standard evaluation coefficient range of the interaction between the pesticide and the crop corresponding to the target crop, it indicates that the efficacy state of the impact of the pesticide corresponding to the target crop on the crop is excessive.

[0056] B4. If the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop is less than the minimum value within the set standard evaluation coefficient range of the interaction between the pesticide and the crop corresponding to the target crop, it indicates that the efficacy state of the impact of the pesticide corresponding to the target crop on the crop is insufficient.

[0057] Step 3. Coordination of sprinkler irrigation and fertilization measures: If the efficacy state of the impact of the pesticide corresponding to the target crop on the crop is excessive or insufficient, an excessive warning prompt and an insufficient warning prompt are given, so as to analyze the level of the impact of the pesticide on the target crop, and then analyze the sprinkler irrigation measures and fertilization measures corresponding to the target crop.

[0058] In a specific embodiment, the process of analyzing the level of the impact of the pesticide on the target crop is as follows: C1. If the efficacy state of the impact of the pesticide corresponding to the target crop on the crop is insufficient, calculate the difference between the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop and the minimum value within the set standard evaluation coefficient range of the interaction between the pesticide and the crop corresponding to the target crop, and obtain the difference between the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop and the minimum value within the set standard evaluation coefficient range of the interaction between the pesticide and the crop corresponding to the target crop. If the difference is greater than or equal to the insufficient threshold, it is recorded as insufficient level 1. If the difference is less than the insufficient threshold, it is recorded as insufficient level 2.

[0059] C2. If the efficacy state of the impact of the pesticide corresponding to the target crop on the crop is excessive, calculate the difference between the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop and the maximum value within the set standard evaluation coefficient range of the interaction between the pesticide and the crop corresponding to the target crop, and obtain the difference between the evaluation coefficient of the interaction between the pesticide corresponding to the target crop and the crop and the maximum value within the set standard evaluation coefficient range of the interaction between the pesticide and the crop corresponding to the target crop. If the difference is less than or equal to the excessive threshold, it is recorded as excessive level 1. If the difference is greater than the excessive threshold, it is recorded as excessive level 2.

[0060] It should be noted that in the link of evaluating the efficacy status and impact level, machine learning is applied. The efficacy status and impact level are evaluated through comparison and difference calculation, and the classification algorithm of machine learning is adopted to improve the accuracy. Data such as the evaluation coefficient of the interaction between pesticides and crops, the interval of the evaluation coefficient of the interaction between pesticides and crops corresponding to the standard target crops, the deficiency threshold, and the overdose threshold are integrated and labeled according to the efficacy status and impact level. Algorithms such as logistic regression, support vector machine, or decision tree classifier are used to train the data, enabling the model to learn the relationship between features and categories. In practical applications, the relevant data of the target crop are input into the trained model, and the model can output the corresponding efficacy status and impact level, providing a scientific basis for the adjustment of subsequent sprinkler irrigation and fertilization measures.

[0061] In another specific embodiment, the sprinkler irrigation measures corresponding to the target crop are analyzed, and the specific analysis process is as follows: D1. If the level of the impact of the pesticides on the target crop is the first level of deficiency, increase the sprinkler irrigation water volume by 10%-15% compared with the conventional sprinkler irrigation volume, ensure that the soil humidity is maintained at 70%-75% of the field capacity, and at the same time, adopt a sprinkler irrigation mode with small water droplets and high frequency, sprinkle once every 2-3 hours, and each sprinkler irrigation lasts for 10-15 minutes. Adjust the sprinkler irrigation time and choose to carry out sprinkler irrigation in the early morning or evening.

[0062] D2. If the level of the impact of the pesticides on the target crop is the second level of deficiency, significantly increase the sprinkler irrigation water volume by 20%-30% compared with the normal sprinkler irrigation volume, ensure that the soil humidity quickly reaches 75%-80% of the field capacity, adopt a combination of flood irrigation and sprinkler irrigation. First, carry out a short period of flood irrigation every 30-45 minutes, and then switch to sprinkler irrigation to maintain the surface soil humidity, sprinkle once every 1 hour. At the same time, add amino acid water-soluble fertilizer with a concentration controlled at 0.2%-0.3%.

[0063] D3. If the level of the impact of the pesticides on the target crop is the first level of overdose, immediately reduce the sprinkler irrigation water volume, reduce the sprinkler irrigation frequency by 30%-40%, and shorten the single sprinkler irrigation time by 20%-30%, so that the soil humidity gradually drops to 60%-65% of the field capacity. Adjust the angle of the sprinkler irrigation nozzle to make it higher than the crop canopy and spray in a mist form.

[0064] D4. If the level of the impact of the pesticides on the target crop is the second level of overdose, suspend sprinkler irrigation for 1-2 days to reduce the soil humidity to 50%-55% of the field capacity. Then, adopt a sprinkler irrigation strategy of small amounts and multiple times, with each sprinkler irrigation volume only 20%-30% of the normal amount, and an interval of 4-6 hours. Add a composite solution containing microbial inoculants to the sprinkler irrigation water with a concentration of 0.1%-0.2%.

[0065] In another specific embodiment, the analysis of the fertilization measures corresponding to the target crop is as follows: E1. If the level of the impact of the pesticide on the target crop is one level short, increase the application rate of nitrogen fertilizer by 10%-15% based on the original planned nitrogen fertilizer application rate. At the same time, apply trace element fertilizers in combination, with a concentration of 0.1%-0.2%, and use foliar spraying. Adjust the fertilization time and frequency, advance the originally planned topdressing time by 3-5 days, and shorten the interval to once every 10-12 days on the basis of originally fertilizing once every two weeks. Adopt a combination of basal fertilization and topdressing. The basal fertilization is mainly organic fertilizer, and the application rate per mu is increased by 5%-10%. The topdressing uses quick-acting chemical fertilizers and is applied by drip irrigation or shallow burial.

[0066] E2. If the level of the impact of the pesticide on the target crop is two levels short, significantly increase the fertilizer application rate. The nitrogen fertilizer application rate is increased by 20%-30% on the original basis. At the same time, apply biological bacterial fertilizer, with a dosage of 10-15 kg per mu. In addition, add fertilizers containing amino acid and humic acid components, with a concentration of 0.3%-0.5% through foliar spraying. Adjust the fertilization time and frequency, shorten the original fertilization interval to 7-10 days, and conduct a one-time centralized fertilization in advance before the high-incidence period of pests and diseases. The fertilization time is selected on cloudy days or evenings. Adopt a three-dimensional fertilization method combining foliar fertilization and root fertilization. For root fertilization, use deep application method to apply the fertilizer to the dense root layer. For foliar fertilization, conduct it once every 5-7 days.

[0067] E3. If the level of the impact of the pesticide on the target crop is one level excessive, reduce the application of nitrogen fertilizer, reduce the original nitrogen fertilizer application rate by 10%-15%, increase the application rate of potassium fertilizer by 10%-15% based on the original planned potassium fertilizer application rate. At the same time, add fertilizers with the function of regulating crop growth. Adjust the fertilization time and frequency, postpone the originally planned topdressing time by 3-5 days, and reduce the fertilization frequency. On the basis of originally fertilizing once every two weeks, extend the interval to once every 16-18 days. Adopt shallow application and scattered application methods, and evenly spread the fertilizer in the shallow soil around the crop roots. For foliar fertilizer, reduce the concentration and application frequency, reduce the foliar fertilizer concentration to 0.05%-0.1%, and adjust the application frequency to once every 10-15 days.

[0068] E4. If the level of the impact of the pesticide on the target crop is two levels excessive, significantly reduce the fertilizer application rate. The nitrogen fertilizer application rate is reduced by 30%-40% on the original basis. At the same time, apply fertilizers with detoxification and repair functions. Adjust the fertilization time and frequency, suspend fertilization for 1-2 weeks, and reduce the fertilization frequency to once a month. Adopt local fertilization method, and open a shallow ditch for fertilization 10-15 cm around the roots.

[0069] In another specific embodiment, the over-dose warning prompt is carried out, and the specific warning process is as follows: F1. If the level of the impact of the target crop pesticide on the crop is the first level of over-dose, a warning prompt is carried out. The warning prompt notification: The title of the APP push notification is "[Crop Name] Pesticide Impact Over-dose First Level Warning", and the notification reads: "Attention! The pesticide effect in your [Crop Name] planting area has reached the first level of over-dose. Please immediately stop the ongoing farming operations, check the mitigation measures recommended by the system, adjust irrigation and fertilization, and reduce the risk of pesticide damage." The plot in the APP is marked with a red warning, and pictures of the possible pesticide damage symptoms of the crop are displayed on the details page to assist the user in judging the on-site situation. The intelligent speaker broadcasts: There is an intelligent speaker installed in the field. The speaker automatically emits an alarm sound, and then broadcasts: "[Crop Name] planting area pesticide over-dose first level, please handle it in time.", and broadcasts continuously for 3 times.

[0070] F2. If the level of the impact of the target crop pesticide on the crop is the second level of over-dose, a warning prompt is carried out. The notification title is "[Crop Name] Pesticide Impact Over-dose Second Level Severe Alarm", and the notification reads: "Very critical! Your [Crop Name] has encountered a second-level over-dose crisis of pesticides, and the crop is already facing a serious threat of pesticide damage. Please organize personnel immediately after receiving the notification and strictly follow the emergency rescue plan given by the system to comprehensively adjust farming activities and do your best to save the crop. If you have any questions, please contact technical support." The plot in the APP will be marked red and flashing, and other irrelevant functions will also be locked, forcing the user to view the rescue plan first. Drone warning: Activate the drones deployed in the field, carry high-power speakers, and circle over the area with the second level of over-dose, playing the alarm sound of "The pesticide in the area below is over-dose at the second level, immediately rescue the crop".

[0071] In another specific embodiment, the lack warning prompt is carried out, and the specific warning process is as follows: G1. If the level of the impact of the target crop pesticide on the crop is the second level of warning, a warning prompt is carried out. The agricultural management system automatically pushes a high-priority notification to the mobile APP of the farmer or the relevant responsible person. The notification title is "[Crop Name] Pesticide Impact Lack First Level Alarm". The farmland management module in the APP will mark the plot in a yellow warning state and highlight it on the map. Click to view the detailed data. At the corresponding field, set up an electronically controlled display powered by solar energy. The screen lights up with a yellow background light and scrolls to display the words "Pesticide lack first level, pay attention to crop growth, refer to the remedial plan".

[0072] G2. If the level of the impact of the pesticide on the target crop is the second-level warning, a warning prompt will be issued. The title of the system push notification will become "[Crop Name] Pesticide Impact Lack Second-level Emergency Alert". In the APP, not only will the plot be marked with an orange warning, but a detailed remedial plan page will also automatically pop up, including step-by-step operation guides and expected effect displays, guiding users to take quick actions. In addition to text messages, an automatic voice call will be arranged to be made to the responsible person's mobile phone, asking the other party to press the "1" key to confirm receipt of the notice. The background light of the field electronic display screen will switch to an orange flashing state, scrolling and displaying "Pesticide Lack Second-level, Act Immediately, Implement Remedy".

[0073] The above content is only an example and illustration of the concept of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the described specific embodiments or use similar methods for substitution, as long as they do not deviate from the concept of the invention or exceed the scope defined in this specification, they should all fall within the protection scope of the present invention.

Claims

1. A method for detecting pesticide spraying based on machine learning, characterized in that: include: Step 1: Analysis of pesticide activity period: After the target crop is sprayed with pesticides, the influencing factors of the pesticide efficacy corresponding to the target crop are obtained, and then the pesticide activity period corresponding to the target crop is analyzed. When the target crop pesticide activity period is reached, several collection points are set to analyze the pesticide-crop interaction evaluation coefficient corresponding to the target crop. Step 2: Assessment of the impact of pesticides on crops: Based on the pesticide-crop interaction assessment coefficient corresponding to the target crop, the efficacy status of the pesticide corresponding to the target crop on the crop is assessed. The efficacy status includes good, excess and deficiency; Step 3: Coordination of sprinkler irrigation and fertilization measures: If the efficacy of the pesticide corresponding to the target crop on the crop is in excess or deficiency, an excess warning prompt and a deficiency warning prompt will be issued to analyze the level of impact of the pesticide on the target crop, and then analyze the sprinkler irrigation and fertilization measures corresponding to the target crop.

2. A method for detecting pesticide spraying based on machine learning as claimed in claim 1, characterized in that: The specific analysis process of analyzing the pesticide activity period corresponding to the target crop is as follows: Obtain the influencing factors of pesticide efficacy corresponding to the target crop, which include crop variety, crop growth stage, pesticide type and pesticide spraying amount. Compare the crop variety, crop growth stage, pesticide type and pesticide spraying amount corresponding to the target crop with the crop variety, crop growth stage, pesticide type and pesticide spraying amount corresponding to each pesticide active period in the database. If the crop variety, crop growth stage, pesticide type and pesticide spraying amount corresponding to the target crop are the same as the crop variety, crop growth stage, pesticide type and pesticide spraying amount corresponding to a certain pesticide active period in the database, then use the pesticide active period in the database as the pesticide active period corresponding to the target crop.

3. A method for detecting pesticide spraying based on machine learning as claimed in claim 2, characterized in that: The pesticide-crop interaction evaluation coefficient corresponding to the target crop is analyzed, and the specific analysis process is as follows: The crop growth assessment value, crop physiological and biochemical assessment value and crop quality assessment value corresponding to the target crop at each collection time point are obtained, and the crop growth assessment value, crop physiological and biochemical assessment value and crop quality assessment value corresponding to the target crop at each collection time point are recorded as , and , where k represents the number corresponding to each acquisition time point, , q is any integer greater than 2 and is represented as the collection of each collection time point, and the crop growth assessment value, crop physiological and biochemical assessment value and crop quality assessment value corresponding to the target crop at each collection time point are substituted into the calculation formula: The pesticide-crop interaction evaluation coefficient corresponding to the target crop is obtained ,in, , and They are the standard crop growth assessment value, standard crop physiological and biochemical assessment value, and standard crop quality assessment value corresponding to the set target crops. , , They are respectively the weight factor corresponding to the set target crop growth assessment value, the weight factor corresponding to the crop physiological and biochemical assessment value, and the weight factor corresponding to the crop quality assessment value.

4. A method for detecting pesticide spraying based on machine learning as claimed in claim 3, characterized in that: The specific acquisition process of obtaining the crop growth assessment value, crop physiological and biochemical assessment value and crop quality assessment value corresponding to the target crop at each collection time point is as follows: A1. Obtain crop growth indicators, crop physiological and biochemical indicators, and crop quality indicators corresponding to the target crops at each collection time point. Crop growth indicators include biomass volume, branch number change rate, and chlorophyll content. Crop physiological and biochemical indicators include photosynthetic pigment content, carboxylation efficiency, and antioxidant enzyme activity change rate. Crop quality indicators include pesticide residues, nutrient content change rate, and shape regularity. A2. Input the biomass volume, branch number change rate and chlorophyll content corresponding to the target crop at each collection time point into the crop growth assessment value analysis model, and output the crop growth assessment value corresponding to the target crop at each collection time point; A3, inputting the photosynthetic pigment content, carboxylation efficiency and antioxidant enzyme activity change rate corresponding to the target crop at each collection time point into the crop physiological and biochemical evaluation value analysis model, and outputting the crop physiological and biochemical evaluation value corresponding to the target crop at each collection time point; A4. Input the pesticide residue, nutrient content change rate and shape regularity corresponding to the target crop at each collection time point into the crop quality assessment value analysis model, and output the crop quality assessment value corresponding to the target crop at each collection time point.

5. A method for detecting pesticide spraying based on machine learning as claimed in claim 4, characterized in that: The specific analysis process of evaluating the efficacy status of the pesticide corresponding to the target crop on the crop is as follows: B1. Compare the pesticide-crop interaction assessment coefficient corresponding to the target crop with the pesticide-crop interaction assessment coefficient interval corresponding to the set standard target crop; B2. If the pesticide-crop interaction assessment coefficient corresponding to the target crop is within the set pesticide-crop interaction assessment coefficient interval corresponding to the standard target crop, it indicates that the efficacy of the pesticide corresponding to the target crop on the crop is good; B3. If the pesticide-crop interaction assessment coefficient corresponding to the target crop is greater than the maximum value within the set pesticide-crop interaction assessment coefficient interval corresponding to the standard target crop, it indicates that the efficacy of the pesticide corresponding to the target crop on the crop is excessive; B4. If the pesticide-crop interaction assessment coefficient corresponding to the target crop is less than the minimum value within the set pesticide-crop interaction assessment coefficient interval corresponding to the standard target crop, it indicates that the efficacy of the pesticide corresponding to the target crop on the crop is deficient.

6. A method for detecting pesticide spraying based on machine learning as claimed in claim 5, characterized in that: The specific analysis process of analyzing the impact of pesticides on target crops is as follows: C1. If the efficacy status of the pesticide corresponding to the target crop on the crop is deficient, the difference between the pesticide-crop interaction assessment coefficient corresponding to the target crop and the minimum value in the set pesticide-crop interaction assessment coefficient interval corresponding to the standard target crop is calculated to obtain the difference between the pesticide-crop interaction assessment coefficient corresponding to the target crop and the minimum value in the set pesticide-crop interaction assessment coefficient interval corresponding to the standard target crop. If the difference is greater than or equal to the deficiency threshold, it is recorded as a first-level deficiency. If the difference is less than the deficiency threshold, it is recorded as a second-level deficiency. C2. If the efficacy state of the pesticide corresponding to the target crop on the crop is excessive, the difference between the pesticide-crop interaction assessment coefficient corresponding to the target crop and the maximum value in the set pesticide-crop interaction assessment coefficient interval corresponding to the standard target crop is calculated to obtain the difference between the pesticide-crop interaction assessment coefficient corresponding to the target crop and the maximum value in the set pesticide-crop interaction assessment coefficient interval corresponding to the standard target crop. If the difference is less than or equal to the excess threshold, it is recorded as the first level of excess; if the difference is greater than the excess threshold, it is recorded as the second level of excess.

7. The method for detecting pesticide spraying based on machine learning as claimed in claim 1, characterized in that: The sprinkler irrigation measures corresponding to the target crops are analyzed, and the specific analysis process is as follows: D1. If the impact of pesticides on target crops is less than level 1, increase the amount of sprinkler irrigation by 10%-15% compared with the conventional amount to ensure that the soil moisture is maintained at 70%-75% of the field water holding capacity. At the same time, use a small droplet, high-frequency sprinkler irrigation mode, sprinkle once every 2-3 hours, and each sprinkler irrigation lasts for 10-15 minutes. Adjust the sprinkler irrigation time and choose to sprinkle in the early morning or evening; D2. If the impact of pesticides on target crops is found to be less than level 2, the amount of sprinkler irrigation should be greatly increased by 20%-30% compared with the normal amount to ensure that the soil moisture quickly reaches 75%-80% of the field water holding capacity. A combination of flood irrigation and sprinkler irrigation should be used. First, a short-term flood irrigation should be carried out every 30-45 minutes, and then sprinkler irrigation should be switched to maintain the surface soil moisture. Sprinkler irrigation should be carried out once every hour. At the same time, amino acid water-soluble fertilizer should be added with a concentration controlled at 0.2%-0.3%; D3. If the impact of pesticides on target crops is analyzed as level one overdose, immediately reduce the amount of sprinkler irrigation, reduce the frequency of sprinkler irrigation by 30%-40%, shorten the single sprinkler irrigation time by 20%-30%, gradually reduce the soil moisture to 60%-65% of the field water holding capacity, adjust the sprinkler nozzle angle to be higher than the crop canopy, and spray in a mist form; D4. If the impact of pesticides on target crops is graded as overdose level 2, suspend irrigation for 1-2 days, reduce soil moisture to 50%-55% of field water holding capacity, and then adopt a strategy of small-scale and multiple irrigations, with each irrigation volume being only 20%-30% of the normal volume, and with an interval of 4-6 hours. Add a compound solution containing microbial agents to the irrigation water at a concentration of 0.1%-0.2%.

8. The method for detecting pesticide spraying based on machine learning as claimed in claim 7, characterized in that: The fertilization measures corresponding to the target crops are analyzed, and the specific analysis process is as follows: E1. If the impact of pesticides on target crops is insufficient for the first level, increase the amount of nitrogen fertilizer applied by 10%-15% on the basis of the original planned amount of nitrogen fertilizer. At the same time, apply trace element fertilizers at a concentration of 0.1%-0.2%. Use foliar spraying to adjust the fertilization time and frequency. Advance the original topdressing time by 3-5 days. On the basis of the original fertilization once every two weeks, shorten the interval to once every 10-12 days. Use a combination of basal fertilizer and topdressing. The basal fertilizer is mainly organic fertilizer, and the application amount per mu is increased by 5%-10%. Use quick-acting fertilizers for topdressing, and apply them by drip irrigation or shallow burial. E2. If the impact of pesticides on target crops is found to be insufficient at level 2, then significantly increase the amount of fertilizers, increase the amount of nitrogen fertilizer by 20%-30% on the original basis, and at the same time, increase the amount of biofertilizers to 10-15 kg per mu. In addition, add fertilizers containing amino acids and humic acid at a concentration of 0.3%-0.5% through foliar spraying, adjust the fertilization time and frequency, shorten the original fertilization interval to 7-10 days, and conduct concentrated fertilization in advance before the high incidence of pests and diseases. The fertilization time should be selected on cloudy days or in the evening. A three-dimensional fertilization method combining foliar fertilization and root fertilization should be adopted. Root fertilization adopts the method of deep application, and the fertilizer is applied to the dense root layer. Foliar fertilization should be conducted every 5-7 days. E3. If the impact of pesticides on target crops is analyzed as level one overdose, reduce the application of nitrogen fertilizer by 10%-15% of the original nitrogen fertilizer dosage, increase the application of potassium fertilizer by 10%-15% on the basis of the original planned potassium fertilizer dosage, and at the same time, add fertilizers that can regulate crop growth, adjust the fertilization time and frequency, postpone the originally planned topdressing time by 3-5 days, reduce the fertilization frequency, extend the interval to once every 16-18 days on the basis of the original fertilization every two weeks, and use shallow and dispersed fertilization to evenly spread the fertilizer in the shallow soil around the crop root system. For foliar fertilizer, reduce the concentration and frequency of use, reduce the concentration of foliar fertilizer to 0.05%-0.1%, and adjust the frequency of use to once every 10-15 days; E4. If the impact of pesticides on target crops is analyzed as level 2 overdose, the amount of fertilizer should be greatly reduced, and the amount of nitrogen fertilizer should be reduced by 30%-40% on the original basis. At the same time, fertilizers with detoxification and repair functions should be applied, and the time and frequency of fertilization should be adjusted. Fertilization should be suspended for 1-2 weeks, and the frequency of fertilization should be reduced to once a month. Local fertilization should be adopted, and shallow trenches should be opened 10-15 cm around the root system for fertilization.

9. A method for detecting pesticide spraying based on machine learning as claimed in claim 8, characterized in that: The specific warning process for over-quantity warning is as follows: F1. If the pesticide impact on the target crop exceeds the level of level 1, an early warning will be issued. The title of the APP push notification is "[Crop Name] Pesticide Impact Exceeding Level 1 Warning". The notification reads: "Attention! The pesticide effect in your [Crop Name] planting area exceeds the level of level 1. Please stop the ongoing farming operations immediately, check the mitigation measures recommended by the system, adjust irrigation and fertilization, and reduce the risk of pesticide damage." The plot in the APP is marked as a red warning, and pictures of possible pesticide damage symptoms of the crop are displayed on the details page to assist users in judging the on-site situation. Smart speaker broadcast: There is a smart speaker installed in the field, and the speaker automatically sounds an alarm, and then broadcasts: "[Crop Name] Pesticides in the planting area exceed the level of level 1. Please deal with it in time", and the broadcast continues for 3 times; F2. If the impact of pesticides on target crops is level 2 overdose, an early warning will be issued with the title "[Crop Name] Pesticide Impact Level 2 Overdose Serious Alarm". The notification reads: "Very critical! Your [Crop Name] is experiencing a level 2 overdose crisis. The crop is facing a serious threat of pesticide damage. Please organize personnel immediately after receiving the notification, strictly follow the emergency rescue plan given by the system, adjust farming activities in all aspects, and do your best to save the crop. If you have any questions, please contact technical support." The plot will be marked in red and flashing in the APP, and other unrelated functions will be locked, forcing users to check the rescue plan first. Drone warning: Enable drones deployed in the field, carry tweeters, hover over the level 2 overdose area, and play the alarm "The area below has level 2 overdose of pesticides, rescue the crop immediately." 10. The method for detecting pesticide spraying based on machine learning according to claim 9, characterized in that: The specific warning process for the shortage warning is as follows: G1. If the impact of pesticides on target crops is at level 2 warning, an early warning will be issued. The agricultural management system will automatically push a high-priority notification to the mobile phone APP of the farmer or relevant responsible person. The notification title is "[Crop Name] Pesticide Impact Lacks Level 1 Alert". The farmland management module in the APP will mark the plot as a yellow warning state and highlight it on the map. Click to view detailed data. In the corresponding fields, a solar-powered electronic display screen will be set up. The screen will light up with a yellow background light and scroll to display the words "Pesticide Lacks Level 1, Pay Attention to Crop Growth, Refer to Remediation Plan"; G2. If the impact of pesticides on target crops is at level 2 warning, a warning will be issued and the title of the system push notification will be changed to "[Crop Name] Pesticide Impact Lack Level 2 Emergency Alert". Not only will the land be marked as an orange warning in the APP, but a detailed remediation plan page will also pop up automatically, including a step-by-step operation guide and a display of expected effects to guide users to act quickly. In addition to text messages, an automatic voice call will be arranged to the responsible person's mobile phone, requiring the other party to press the "1" key to confirm receipt of the notification. The background light of the electronic display screen in the field will switch to an orange flashing state, and the message "Pesticide Lack Level 2, Take Immediate Action, Implement Remediation" will be displayed in a scrolling manner.

Images