Tobacco field variable rate fertilization control system and method

Abstract

The invention relates to the technical field of self-adaptive control, in particular to a tobacco field variable fertilization control system and method, and the system comprises a sensing data collection module which collects soil humidity data through a soil humidity sensor, monitors the growth data of crops at the same time, and carries out the synthesis to obtain real-time soil and growth data; and performing time sequence analysis on the real-time soil and growth data, identifying a data trend and an abnormal point, and generating a trend analysis result. According to the method, the soil humidity and the crop growth data are collected, and the time sequence analysis is combined, so that the trend and the abnormality are quickly identified, and the prediction accuracy is improved. The convolutional neural network is used for carrying out recognition and probability evaluation on diseases and insect pests, so that disease and insect pest management is more accurate, and response is quicker. Through the technologies, the system can dynamically adjust irrigation parameters based on real-time data, optimize water distribution, and ensure that an irrigation strategy is perfectly matched with crop requirements and potential risks.

Description

Technical Field

[0001] The invention relates to the technical field of adaptive control, and in particular to a tobacco field variable fertilization control system and method. Background Art

[0002] Adaptive control technology is a control system technology that enables the system to automatically adjust its control parameters when the external environment or system parameters change to maintain or improve system performance. The tobacco field variable fertilizer control system is an intelligent control system designed specifically for the tobacco planting field. Its main purpose is to automatically adjust the amount of fertilizer or irrigation to adapt to the changing soil and environmental conditions in the tobacco field.

[0003] However, existing technologies rely on regular testing and manual intervention, which results in a slow response and makes it difficult to effectively respond to short-term environmental changes or pest and disease emergencies. This processing lag may not only lead to a waste of resources, but also increase the risk of crop diseases due to improper irrigation. In addition, existing technologies lack highly integrated data analysis capabilities, making it difficult to achieve refined management, thus affecting the overall efficiency of crop production and environmental sustainability. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a tobacco field variable fertilization control system and method.

[0005] In order to achieve the above object, the present invention adopts the following technical scheme: A variable fertilization control system for tobacco fields comprises: The sensing data acquisition module collects soil moisture data through a soil moisture sensor and monitors crop growth data to synthesize real-time soil and growth data; performs time series analysis on the real-time soil and growth data to identify data trends and abnormal points and generate trend analysis results; a pest and disease analysis module, receiving the trend analysis result, analyzing the pest and disease signs in the crop image through image recognition and convolutional neural network, identifying the pest and disease type, and generating a pest and disease identification result; evaluating the probability of pest and disease occurrence according to the pest and disease identification result, and generating a pest and disease occurrence probability evaluation result; The irrigation control module adjusts the irrigation amount according to the pest occurrence probability assessment result, optimizes the water pump rate and switching time to match the current crop demand and the predicted pest risk, and generates adjusted irrigation parameters; combines the matching degree of the adjusted irrigation parameters with the real-time soil moisture data, optimizes water distribution, and obtains the optimal irrigation configuration; The fertilization decision module is based on the irrigation optimization configuration and collects real-time data of soil moisture to determine the need for chemical fertilizer adjustment, and selects the type and application amount of chemical fertilizer based on the growth requirements of crops and the chemical properties of the soil.

[0006] Preferably, the steps for acquiring the real-time soil and growth data are: Soil moisture sensors are used to monitor soil moisture in tobacco fields, and cameras are used to capture crop growth status to obtain preliminary soil moisture data and crop growth data. Based on the preliminary soil moisture data and crop growth data, aligning and merging time stamps to form a unified data set; Based on the unified data set, the interaction between soil moisture and crop growth is analyzed, and real-time soil and growth data are generated through fusion processing.

[0007] Preferably, the steps of obtaining the trend analysis results are: Based on the real-time soil and growth data, statistical analysis is performed, including calculation of mean, standard deviation and extreme value, to obtain statistical characteristic analysis results of the data; Based on the statistical characteristics analysis results of the data, time series analysis is performed to calculate the comprehensive score at each time point. The calculation formula is: ; in, is the comprehensive score at time t, is the observed value at time t, and are the mean and standard deviation of the data set, respectively. is the moving average, is the standard deviation of the moving average, and is the weight coefficient; Based on the comprehensive score, the overall trend is evaluated and abnormal points are identified to obtain trend analysis results.

[0008] Preferably, the steps for obtaining the pest identification result are: Using the trend analysis results, selecting crop images for pest and disease sign analysis to obtain selected crop images; Based on the selected crop image, deploy a convolutional neural network to perform image processing, identify pest and disease characteristics in the image, and obtain a pest and disease characteristic analysis result in the image; Based on the analysis results of the pest and disease characteristics in the image, the pest and disease types are classified, and by comparing with an established pest and disease characteristic database, the pest and disease types are identified and calibrated to obtain pest and disease identification results.

[0009] Preferably, the steps for obtaining the pest occurrence probability assessment result are: Receiving the pest identification result, including the type of each pest and the distribution characteristics in the image, and obtaining pest distribution data; Based on the pest distribution data, the occurrence probability of various pests is calculated, and the calculation formula is: ; in, is the probability of occurrence of the i-th pest, Represents the coverage area of ​​pests and diseases in the identification results, Represents the degree of impact, represents temperature sensitivity, Represents seasonality, is the adjustment coefficient, n is the total number of pests and diseases; Based on the occurrence probability, a risk assessment is conducted, and combined with environmental factors and past data, an assessment result of the occurrence probability of pests and diseases is obtained.

[0010] Preferably, the steps for obtaining the adjusted irrigation parameters are: Analyze the probability of each type of pests and diseases provided by the pests and diseases occurrence probability assessment results, and conduct an impact analysis on irrigation demand based on the probability, including investigation of crop growth cycles and water requirements, to determine the range of initial irrigation volume to be adjusted; Based on the preliminary irrigation volume range, calculate the pump rate using the formula: ; in, is the adjusted pump speed, is the basic pump speed, is the amount of water currently required by crops, is the standard water volume, is the pest risk adjustment factor; Based on the water pump rate, the irrigation parameters are reconfigured to obtain adjusted irrigation parameters.

[0011] Preferably, the steps for obtaining the irrigation optimization configuration are: Based on the adjusted irrigation parameters and soil moisture data, the matching index between the two is calculated, and the calculation formula is: ; in, is the matching index, is the ideal soil moisture at the i-th detection point, is the soil moisture at the i-th detection point, is the number of detection points; The operating parameters of the water pump and the irrigation plan are optimized according to the matching index to obtain an optimal irrigation configuration.

[0012] The present invention provides a tobacco field variable fertilization control method, comprising the following steps: Soil moisture data is collected through soil moisture sensors, and crop growth data is collected at the same time. Time series analysis is performed on the collected data to identify data trends and anomalies, and obtain comprehensive analysis results of soil and growth data; Based on the comprehensive analysis results of soil and growth data, image recognition and convolutional neural networks are used to analyze the signs of pests and diseases in crop images, identify the types of pests and diseases, assess the probability of pest and disease occurrence, and generate pest and disease identification and risk assessment results; Based on the results of pest and disease identification and risk assessment, adjust the irrigation amount, optimize the pump rate and switching time to match the current crop demand and predicted pest and disease risks, and generate adjusted irrigation parameters; Combine the adjusted irrigation parameters with the real-time soil moisture data to optimize water distribution and obtain the optimal irrigation configuration result; Using the irrigation optimization configuration results, the type and application amount of chemical fertilizers are selected according to the growth requirements of crops and the chemical properties of the soil.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, soil moisture and crop growth data are collected and combined with time series analysis to achieve rapid identification of trends and anomalies, thereby improving the accuracy of predictions. Convolutional neural networks are used to identify and probability evaluate pests and diseases, making pest and disease management more accurate and responsive. Through these technologies, the system can dynamically adjust irrigation parameters based on real-time data, optimize water distribution, and ensure that irrigation strategies are perfectly matched to crop needs and potential risks. This method not only reduces resource waste, but also effectively reduces the potential impact of pests and diseases, promotes healthy growth of crops, and improves yield and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] See also Figure 1 The present invention provides a technical solution: a variable fertilization control system for tobacco fields comprises: The sensing data acquisition module collects soil moisture data through soil moisture sensors and monitors crop growth data at the same time, synthesizing real-time soil and growth data; performs time series analysis on real-time soil and growth data, identifies data trends and anomalies, and generates trend analysis results; The pest and disease analysis module receives the trend analysis results, analyzes the pest and disease signs in the crop images through image recognition and convolutional neural network, identifies the pest and disease types, and generates pest and disease identification results; evaluates the probability of pest and disease occurrence based on the pest and disease identification results, and generates pest and disease occurrence probability evaluation results; The irrigation control module adjusts the irrigation volume according to the probability assessment results of pests and diseases, optimizes the pump rate and switching time to match the current crop demand and the predicted pest and disease risk, and generates adjusted irrigation parameters; combines the adjusted irrigation parameters with the matching degree of real-time soil moisture data to optimize water distribution and obtain the optimal irrigation configuration; The fertilization decision module is based on the irrigation optimization configuration and collects real-time soil moisture data to determine the need for chemical fertilizer adjustment. It also selects the type and amount of chemical fertilizer based on the growth requirements of crops and the chemical properties of the soil. Specifically, based on the irrigation optimization configuration, the real-time data of soil moisture is collected at the same time. First, the adjusted irrigation parameters are read and the corresponding specific irrigation period, irrigation amount and other information are arranged in parallel with the real-time records of soil moisture. For each soil moisture collection point, its geographical location and sensor number are checked. Based on the factory calibration value of the sensor, its current output is checked and whether there is a transmission delay. If a sensor output is found to be unstable, the item is recorded and its line or power connection is checked again. After confirming that all sensor outputs are within the normal range, the multiple soil moisture readings obtained are compared with the current growth stage of the crop. The growth stage can be calculated through the sowing time to get the initial stage, elongation stage or flowering and fruiting stage and confirmed in combination with the height and number of leaves of the plant. After summarizing the soil moisture and crop growth stage information, the soil chemical property parameters of the field are retrieved. The chemical property parameters are generally obtained through regular field sampling and measured in the agricultural laboratory, including pH value, organic matter content and various nutrient indicators, and then the reference crop nitrogen, phosphorus, Under the guidance of the absorption ratio of chemical elements such as potassium and the field fertilization specifications, the fertilizer recommendation table accumulated locally is retrieved respectively. The fertilizer recommendation table includes common types of chemical fertilizers and their applicable crop growth stages. The proportion of active ingredients and release duration of each chemical fertilizer are obtained by agricultural technicians through summarizing the test data of previous experiments. Then, the simultaneous absorption of water and nutrients by crop roots is analyzed in combination with the collected real-time data of soil moisture. If it is found that the soil moisture in some areas has long deviated from the threshold set by years of experience, the types of chemical fertilizers and application methods in those areas will be readjusted. For example, the original whole bag application will be changed to batch application in multiple time periods, or the ratio of granular and liquid fertilizers will be adjusted. After completing these targeted analyses, the required types of chemical fertilizers and their respective application amounts are obtained, and the application amount is recorded as kilograms per hectare or grams per crop. It is matched with the irrigation optimization configuration obtained earlier, and the results are summarized at each key stage of the crop growth cycle to form a final adjustment plan, and finally the type and application amount of chemical fertilizers are determined.

[0017] The steps to obtain real-time soil and growth data are: Soil moisture sensors are used to monitor soil moisture in tobacco fields, and cameras are used to capture crop growth status to obtain preliminary soil moisture data and crop growth data. Based on the preliminary soil moisture data and crop growth data, the time stamps are aligned and merged to form a unified data set; Based on a unified data set, the interaction between soil moisture and crop growth is analyzed, and real-time soil and growth data are generated through fusion processing.

[0018] Specifically, multiple soil moisture sensors are deployed in the tobacco field area and placed at different depths. Each sensor is first calibrated initially. During calibration, several reference values ​​that meet local planting conditions are selected. The reference values ​​are derived from historical planting experience and multiple measured averages of similar soil environments. If the monitoring data of a sensor deviates from the reference value by more than the agreed comparison value, the sensor installation position and line connection are rechecked while recording the data. After calibration, the camera is installed in a fixed position that can cover the target crop growth area and the appropriate resolution and shooting frequency are set. The camera startup time and the time when the soil moisture sensor starts monitoring are recorded as a basis for subsequent correlation. Then, the soil moisture changes and the growth images of the crops in the picture are continuously acquired and digitally encoded, and finally preliminary soil moisture data and crop growth data are obtained.

[0019] Based on the preliminary soil moisture data and crop growth data obtained earlier, the time tags of the sensor output and the time tags of the camera shooting frames are selected for pairing and alignment. For the parts with inconsistent timestamps, the inserted or discarded time periods are first confirmed by comparing the values ​​of adjacent moments with the image frame numbers, and the alignment details are recorded in the comparison table. The comparison table is generated by the actual monitoring start time, sensor output frequency and camera shooting frequency. If missing records or picture information are encountered, they are filled by interpolation or repeated reading. The interpolation threshold is determined by the minimum update interval under the observation conditions and calibrated by manual experience. If data segments that cannot be aligned normally still appear after filling, they are manually marked and the original data are retained for subsequent inspection. After all time points are aligned and merged, a unified data set is formed.

[0020] Based on the unified data set, the soil moisture readings in each record and the growth indicators such as crop height and leaf number in the corresponding crop screen are first extracted, and these indicators are matched in the same row of records according to the reference arrangement order at the same time, and the linear regression algorithm is used to measure the correlation between soil moisture and each growth indicator. The linear regression algorithm splits the data set by selecting the training ratio and the verification ratio. For example, 70% of the data items are selected as the training subset and 30% of the data items are selected as the verification subset. In the training subset, the independent variable of the regression equation is set as the soil moisture value and the dependent variable is a growth indicator of the target crop. The corresponding regression coefficient is iteratively fitted using the least squares method, and then the regression coefficient is applied to each sample in the verification subset to check the output error. If the error exceeds the preset limit obtained by statistics on the planting data of previous years, the bias term of the regression equation is readjusted and iterated continuously. A relatively stable regression model is obtained through multiple corrections. Subsequently, the model is synchronously input into the fusion module with the real-time collected soil moisture readings and crop growth indicators. When receiving new data, the fusion module calls the trained regression equation for a one-time calculation and outputs real-time soil and growth data, and finally generates real-time soil and growth data.

[0021] The steps to obtain trend analysis results are: Based on real-time soil and growth data, statistical analysis is performed, including calculation of mean, standard deviation and extreme value, to obtain statistical characteristic analysis results of the data; Based on the statistical characteristics analysis results of the data, time series analysis is performed to calculate the comprehensive score at each time point. The calculation formula is: ; in, is the comprehensive score at time t, is the observed value at time t, and are the mean and standard deviation of the data set, respectively. is the moving average, is the standard deviation of the moving average, and is the weight coefficient; Based on the comprehensive score, the overall trend is evaluated and outliers are identified to obtain trend analysis results.

[0022] Specifically, based on the real-time soil and growth data, all records are first organized into a numerical sequence, then accumulated one by one and the accumulated sum is divided by the total number of records to obtain the mean, where the total number of records can be confirmed by counting the collected soil and growth data item by item, and then the difference between each record and the mean is calculated and squared and accumulated. The square accumulated value is divided by the total number of records and the square root is taken to obtain the standard deviation. Then, the maximum and minimum record points are selected in the entire numerical sequence to determine the extreme values. Finally, the statistical analysis of the real-time soil and growth data is completed to obtain the statistical characteristic analysis results of the data.

[0023] The benefit of the formula is that it introduces two sets of differences and weights them separately to obtain a comprehensive measurement value. and Used to balance the impact of different differences on the comprehensive score. and It can be obtained by calculating the mean and standard deviation of the previously obtained data series. For example, from 500 sets of soil moisture and growth parameters collected continuously, and , The measured value at time t can be recorded as 50, and It is obtained by accumulating and calculating one by one through a moving window. For example, the weighted average of the last five pieces of data is used to obtain And its standard deviation ,Pick and , then substituting these data into the formula in turn, the multi-level operation process can be obtained as follows: First calculate , and then , then order , then use and Weighted , we can get The result shows that the comprehensive score at this moment is about 0.86. When the score is greater than a certain empirical control value, it can be regarded as abnormal or significantly deviated. When it is lower than the control value, it is in a relatively stable range.

[0024] Based on the comprehensive score, the comprehensive score curve within a period of time is selected and matched with the observation time. First, a threshold range from historical analysis is set. The range is determined by the statistical interval of previous data and the expert experience in the actual monitoring environment. If the comprehensive score continues to be higher than the threshold range for multiple consecutive moments, the corresponding soil moisture or growth index will be included in the key focus list while recording these moments, and each record will be tracked. If it is found that the comprehensive score falls below the threshold range within a certain period of time, the subsequent time period will be observed and the original recording frequency will be maintained. When the comprehensive scores of all moments are compared and marked, the trend analysis results are obtained.

[0025] The steps to obtain pest and disease identification results are as follows: Using the trend analysis results, crop images are selected for pest and disease sign analysis to obtain selected crop images; Based on the selected crop images, a convolutional neural network is deployed to process the images, identify the characteristics of pests and diseases in the images, and obtain the analysis results of the characteristics of pests and diseases in the images; Based on the analysis results of pest and disease characteristics in the image, the pest and disease types are classified. By comparing with the established pest and disease feature database, the pest and disease types are identified and calibrated to obtain the pest and disease recognition results.

[0026] Specifically, the crop images are screened according to the trend analysis results obtained above. First, the corresponding date and shooting number are locked in the image database according to the time segment indicated in the trend analysis results. After all the images under these dates are collected, their shooting focal lengths and lighting conditions are detected. If it is found that some images cannot be identified due to overexposure or perspective offset, they are first eliminated. The remaining images are further grouped according to the area ratio of the crop in the picture. The area ratio is calculated by reading the proportion of green pixels in the image and combining the crop height information. The threshold is set by daily monitoring experience combined with the average plant height of local crops. If it is detected that the proportion of green pixels in some images is too low and the position of the main crop deviates from the center by more than the defined value given by the expert's actual measurement, the image is excluded from the subsequent analysis queue. After obtaining images that meet the quality requirements, these images are preliminarily judged for signs of pests and diseases. The judgment basis includes features such as whether there are curled spots on the crop leaves and whether there are abnormal lines on the stems. When the above signs are detected, the image is recorded and the pixel coordinate range of the suspicious area is marked. After multiple records and comparisons are confirmed, the final screening is completed and the crop images that need to be deeply analyzed this time are obtained, and the selected crop images are finally obtained.

[0027] Deep processing is performed on the selected crop images. First, the corresponding multi-channel data input form is assigned to each image and the row and column dimensions are specified. Each pixel is numerically normalized on the row and column dimensions. At the same time, the background parts of the image that are not related to the crop body are removed through the set color intervals. These color intervals are set by analyzing the pixel statistics of several background images. If there are cases where the pixel mean and chromaticity distribution are significantly deviated from the crop body, its pixel matrix is ​​marked as an invalid area. After the image data is preprocessed, a convolutional neural network is deployed to train it. The network contains several convolutional layers and pooling layers. The recognition accuracy is enhanced by extracting the texture features formed on the surface of leaves, fruits and stems layer by layer. The network weights are obtained by training with images of healthy crops and common pests and diseases collected in the early stage. During training, forward operations are performed on each batch of image samples and the output is compared with the known labels. The weights are iteratively updated through error back propagation. After multiple rounds of training, these weights are fixed to form the final model. Then, the selected crop images are convolved one by one in the inference mode and the feature response matrix is ​​output. These response matrices are then classified and analyzed, and the location and significance of the suspected pest and disease areas are marked, finally obtaining the pest and disease feature analysis results in the image.

[0028] After obtaining the analysis results of pest and disease features in the image, each pest and disease feature is classified into types. First, the key attribute indicators of pests and diseases collected from a reference are called, and these key attribute indicators are broken down into multiple dimensions such as appearance, color distribution, and grain shape. Then, the pest and disease features in the current image are compared separately. If the degree of match with a certain pest and disease type is high at the morphology and grain level, the pest and disease feature is marked as this type. The degree of match is determined by the query results of the established pest and disease feature database, which contains tens of thousands of image feature vectors and known pest and disease labels. Multiple fields are used for item-by-item comparison during retrieval, including spot shape indicators, degree of leaf defect, etc. If multiple types coexist, the image is included in the composite type statistics at the same time. Through such step-by-step classification, the actual pest and disease types are identified and calibrated in turn, and finally the pest and disease recognition results are obtained.

[0029] The steps to obtain the results of the probability assessment of pests and diseases are as follows: Receive pest and disease identification results, including the type of each pest and disease and its distribution characteristics in the image, and obtain pest and disease distribution data; Based on the pest distribution data, the occurrence probability of various pests and diseases is calculated using the following formula: ; in, is the probability of occurrence of the i-th pest, Represents the coverage area of ​​pests and diseases in the identification results, Represents the degree of impact, represents temperature sensitivity, Represents seasonality, is the adjustment coefficient, n is the total number of pests and diseases; Based on the probability of occurrence, risk assessment is carried out, and the probability assessment results of pest and disease occurrence are obtained by combining environmental factors and past data.

[0030] Specifically, the pest and disease identification results are received and all types and their distribution characteristics in the image are obtained by category. First, according to the pest and disease type information indicated in the identification results and the corresponding coverage range, the corresponding pixel coordinate areas in the image are marked one by one, and the position and area range of each identified pest and disease in each image are archived and the relative position relationship between them and the crop characteristics is recorded. Then, the repeated distribution positions of the same pest and disease type are merged and sorted. In this process, if it is found that the overlapping area of ​​the boundaries of the same pest and disease type is repeatedly marked exceeds the judgment value specified based on empirical investigation (for example, 10% of the pixel coverage area), these overlapping areas are merged into a single The distribution area of ​​the pests and diseases is determined and its area parameters are updated. When the fit between the distribution area coordinates and the crop contour is lower than the comparison threshold specified by some expert experience (for example, 60% fit), the area is marked as a suspicious area and the adjacent frame images are read again to confirm whether there is any recognition deviation. If the recognition is normal after proofreading multiple frames of images, the coordinates and area information of the pest and disease distribution area are retained, and all the proofread and merged pest and disease distribution areas are classified and counted according to the types of pests and diseases and their distribution. The number of occurrences and the total coverage area of ​​each pest and disease need to be recorded separately in the statistics to avoid confusion between the area and the number of occurrences in the later stage. Finally, these statistical data are summarized to form complete pest and disease distribution data.

[0031] The formula is useful in that it integrates four indicators: pest coverage area, impact degree, temperature sensitivity, and seasonality. It is normalized and calculated in exponential form, so that the impact factors of the same or different orders of magnitude can be reasonably quantified. The steps for obtaining each parameter are as follows: coverage area The influence degree is obtained by matching the pixel area in the image recognition result with the field measurement data. The agricultural technicians quantified the yield loss caused by pests and diseases and recorded it as a score between 0 and 10. Temperature sensitivity It is derived from the temperature threshold corresponding to the local growth period of pests and diseases and converted into a range of 0 to 5. Obtained through monthly or quarterly statistical analysis, it indicates the reproduction intensity of pests and diseases in different seasons and is expressed with a score of 0 to 5. The adjustment coefficient The parameters in previous years are revised step by step according to their influence on the final probability results, and their fluctuation range in the actual environment is recorded. Calculation process: Take the total number of pests and diseases , of which the coverage area of ​​the first pest is (Unit is square centimeters, obtained by combining the identification results and field measurements), impact degree , temperature sensitivity , seasonal , and set , substituting these values ​​into the exponential terms, we get: ; Multiplying and adding each term gives: ; Then calculate , the same numerical input and index calculation are performed for other pests and diseases (type 2 and type 3), and the obtained and Two results, add the three together as the denominator, and divide them by the denominator to get The probability of occurrence of three pests and diseases; The results show the probability value corresponding to each pest and disease under the current environmental parameters and identification data. If the probability value is higher than the reference boundary established through data from previous years (for example, 0.4), the subsequent processing should focus on the risk of its spread in a specific plot. If it is lower than this boundary, it is regarded as a general occurrence probability range.

[0032] Based on the above-mentioned probability of occurrence, when conducting risk assessment, the probability of each pest and disease is first compared with the local field monitoring data in the past three to five years. These historical data include annual average temperature, precipitation, soil pH, and crop population density information. In the comparison stage, the types of pests and diseases and their coverage in different months of each year are loaded one by one, and this information is compared with the latest calculated probability of occurrence. If the actual number of past occurrences of certain pests and diseases shows a significant increase or decrease under similar climatic conditions, the difference will be included in the assessment and the score will be adjusted. The adjustment of the score is based on the reference through The statistical standards obtained through regular consultations by local agricultural technical departments, such as setting an additional risk score for pests and diseases caused by a 1°C increase in temperature, or certain pest and disease derivative factors under conditions of continued low precipitation, are used to construct a unified risk level system after adjusting the pest and disease probabilities. Each pest and disease category is matched with its comprehensive risk value. If the comprehensive risk value exceeds the boundary set by the expert group, it is marked as high risk; if it falls within the normal range, it is marked as regular risk. Finally, the risk level of each pest and disease and the corresponding probability of occurrence are summarized and processed to obtain the pest and disease occurrence probability assessment result.

[0033] The steps to obtain the adjusted irrigation parameters are: Analyze the probability of each type of pest and disease provided by the pest and disease occurrence probability assessment results, and conduct an impact analysis on irrigation demand based on the probability, including investigation of crop growth cycle and water demand, to determine the initial irrigation amount range for adjustment; Based on the preliminary irrigation volume range, calculate the pump rate using the following formula: ; in, is the adjusted pump speed, is the basic pump speed, is the amount of water currently required by crops, is the standard water volume, is the pest risk adjustment factor; Based on the water pump rate, the irrigation parameters are reconfigured to obtain adjusted irrigation parameters.

[0034] Specifically, the probability of each type of pest and disease provided by the pest and disease occurrence probability assessment results is analyzed. First, the probability value of each pest and disease is read from the assessment result and compared with the recent growth stage of the crop. The growth stage information is obtained through the key nodes such as crop sowing time, seedling stage to heading stage and corresponding physiological characteristics survey recorded in the planting archives. Then, according to the sensitivity of crops to water demand at different stages, the corresponding irrigation time and water consumption references are listed one by one. For the stage with a higher probability of pests and diseases, the relationship between irrigation amount and irrigation period is focused on. If the irrigation demand statistics for similar crops in the planting habits of the region show that low water volume in a certain period of time is likely to cause the spread of pests and diseases, then the irrigation amount for this period will be increased in combination with the crop water demand model accumulated by agricultural technicians. Conversely, if the probability of pests and diseases in a certain period of time is found to be low, the irrigation amount can be reduced until it is maintained. To ensure that crops absorb and maintain soil moisture normally above the standard threshold set by the agricultural technology department, this standard threshold is based on field monitoring and statistics of the local average precipitation over many years, crop root depth and soil water holding capacity. After completing the comparison of the relationship between the probability of pests and diseases and water demand in all stages, the water demand adjustment situation in each period is obtained and combined into a preliminary irrigation volume range. If the irrigation volume in some high-pest and disease risk stages is planned to be too high, it is necessary to verify again whether it exceeds the safety upper limit obtained from the actual irrigation conditions in multiple consecutive seasons. The safety upper limit is also summarized by the agricultural technology department through years of accumulated field observations. If there is no over-limit situation, the proposed amount can be maintained. If there is an over-limit, the period will be appropriately adjusted down without affecting crop absorption. Finally, after confirming the minimum, regular and maximum irrigation volume ranges for each period, a complete preliminary irrigation volume range is formed.

[0035] The benefit of the formula is that by multiplying the square root term and the risk coefficient increment term, the actual water requirement of the crop is compared with the standard water requirement, and then appropriately enlarged or reduced according to the risk of pests and diseases, so that the overall irrigation efficiency is more in line with the current field conditions. represents the adjusted pump speed, represents the basic water pump rate measured under normal environmental and soil moisture conditions, The current monitored crop water requirement is obtained by measuring the moisture content at the surface and root depth of the field and combining it with the survey results of the water absorption characteristics of the crop varieties. It is based on the standard amount of water required for each irrigation cycle of general crops calculated from local planting records over many years. It comes from the previous pest and disease occurrence probability assessment results, which represents the corresponding pest and disease risk adjustment coefficient on the percentage scale; Calculation process: Through previous measurements and statistics in the same area, (Unit: L / min), (Unit L), (unit L), and refer to the pest risk assessment , substitute the above data item by item, and first calculate the square root part: ; Then calculate the increment term: ; Finally, we multiply the three terms together: ; After multiple levels of operations, we can get (Unit: L / min); The results show that in this example, the new water pump rate is increased to about 19.71L / min. If this value is further greater than the water pump safety upper limit set by the local agricultural technology department, it should be reviewed in actual operation. If it is within a reasonable range, this value can be put into use directly. The larger the specific value, the more water output in the same time period, which needs to be monitored and adjusted in combination with soil infiltration capacity and drainage measures.

[0036] After obtaining the adjusted water pump rate, the irrigation parameters are reconfigured. First, the new water pump rate is compared with the terrain slope, pipeline distribution and drainage path of the field where the crop is located, and the corresponding relationship between these factors and the water pump rate is recorded respectively. For example, in fields with larger slopes, the water supply pressure head needs to be increased accordingly to maintain a stable flow rate. The farthest water supply distance between each area and the water pump interface is recorded. If it is found that the water flow rate at the end of some pipes has dropped significantly, the field should be set as a priority adjustment unit and additional water supply time should be allocated. For the drainage path, it is also necessary to confirm point by point whether it can carry the new flow. If the local drainage is insufficient, the channel should be dug or dredged or modified. After completing the adaptation and adjustment of the pipeline information and water pump rate of each field, the fields are irrigated quantitatively in stages, and the actual time consumed in the current irrigation round and the trend of soil moisture changes are recorded. If the soil moisture detection results are within the range determined by agricultural technicians and the potential risk of pests and diseases has not changed significantly, the irrigation parameters of this time can be continued to be applied to the subsequent stages. On the contrary, if it is monitored that the soil moisture after irrigation deviates seriously from the normal value or the level of pests and diseases increases abnormally, it is necessary to correct the water pump rate again in combination with the risk adjustment coefficient. Finally, the irrigation plan for all subdivided areas is confirmed and the corresponding irrigation period, rate and duration are output to obtain the adjusted irrigation parameters.

[0037] The steps to obtain the optimal irrigation configuration are: Based on the adjusted irrigation parameters and soil moisture data, the matching index between the two is calculated using the following formula: ; in, is the matching index, is the ideal soil moisture at the i-th detection point, is the soil moisture at the i-th detection point, is the number of detection points; The operating parameters of the water pump and the irrigation plan are optimized according to the matching index to obtain the optimal irrigation configuration.

[0038] Specifically, the formula is beneficial in that the difference between the ideal soil moisture and the actual soil moisture at each detection point is normalized and accumulated, and then mapped to an interval between 0 and 1 in an exponential form, thereby summarizing the degree of deviation between the irrigation parameters and the soil moisture data, where The closer the value is to 1, the closer the two are, and the closer it is to 0, the greater the deviation; It represents the ideal soil moisture at the i-th detection point. This value is taken from the optimal range obtained by long-term observation of the growth habits of crops planted in the field and the local climate (for example, some vegetables may be suitable for soil moisture between 20% and 30%). In actual use, it can be further subdivided according to different crop types and growth stages. The soil moisture sensor collects data in real time and obtains the value by combining the depth position of each detection point and the sensor calibration parameters. For the number of detection points, multiple detection locations can be set according to the field area and micro-topography differences; Calculation process: Deployed in a certain plot Detection points to obtain ideal humidity The actual humidity is 25, 26, 28, 28, 30 (in percentage). They are 24, 25, 30, 29, and 31 respectively (the unit is also percentage), and the deviation term corresponding to the i-th detection point is , we can calculate item by item: ; ; ; ; ; Adding these deviations gives: ; Recalculate the exponential term , so we have: ; The results show that the matching index is about 0.55, indicating that there is a certain deviation between the current irrigation parameters and the soil moisture data. When the value is greater than 0.8, it can be regarded as a higher match, and when it is less than 0.4, it can be regarded as a lower match. The intermediate range is further analyzed according to the intervals subdivided by the agricultural technology department.

[0039] When analyzing the parameters in the matching index, first collect the information of all detection point locations and check their coordinate distribution in the field. Record the ideal humidity source of each detection point as a column of values ​​and mark that the ideal humidity comes from the optimal moisture content of the crop in the current growth stage. Then, when reading the actual humidity, it is necessary to focus on checking the stability of the sensor output. In the calibration process, adjust the benchmark value of each sensor. The benchmark value is usually obtained by measuring the control sample. When it is determined that all sensor outputs are consistent, compare the current multiple detection point records with the ideal humidity, calculate the relative deviation, and separately count the single-point deviation and the comprehensive deviation. If it is found that the deviation of individual detection points has been higher than the threshold formed by years of monitoring accumulation (for example, 0.25), record the detection point and check or replace the line again. At the same time, check the local terrain, drainage ditch patency and crop in the field. The root zone depth is checked to ensure that there is no local ground hardening or blockage that affects the data collection of the sensor. When it is confirmed that the detection error is within the normal range, all deviation values ​​are added one by one and converted into an index to obtain the matching index of the entire field. Then the matching index is compared with the previously determined irrigation parameters. If the matching index continues to be low, it is checked whether there are extreme temperatures, heavy rainfall or other abnormal growth conditions of crops during this period, and the irrigation water volume and irrigation time of the next stage are adjusted according to the specific situation. Finally, these processes are summarized into a systematic record to facilitate historical tracing when calling soil moisture data in subsequent stages, complete the verification of the matching index and the analysis of the degree of consistency with the adjusted irrigation parameters, thereby saving the ideal humidity, actual humidity and the corresponding difference results in the same comparison to obtain more comprehensive soil moisture dynamic data.

[0040] Based on the matching index obtained above, when further optimizing the operating parameters of the water pump and the irrigation plan, it is necessary to first extract the changing trend of the matching index on different dates or time periods, list the differences in the high temperature period, precipitation period or night period one by one, and arrange them in correspondence with the previously configured irrigation time, irrigation flow, etc. If it is found that the matching index increases significantly at night or in rainy weather, it may mean that the existing irrigation parameters are redundant in certain periods, or it may indicate that the soil has better water retention in these periods. At this time, the operating power of the water pump is fine-tuned and the power is transferred to a smaller range to avoid excess water waste and ensure uniform humidity distribution between detection points. Check the order and duration of irrigation rotation for each field in the irrigation plan. If the matching index of some fields is high for a long time, it means that they are in a state closer to the ideal humidity and the water supply period can be appropriately shortened. If the matching index of some fields is found to drop continuously in drought weather, it means that they need more precise irrigation scheduling. By comparing the meteorological records of previous years and the actual needs of crops, increase the water supply period of the field in the existing plan or increase its water pump output rate. When the matching index of all fields is uniformly updated and recorded again, these updated values ​​are included in the subsequent irrigation data statistics, and finally the irrigation optimization configuration is generated and compared with the previous plan for reference.

Claims

1. A variable fertilization control system for tobacco fields, characterized in that: The system comprises: The sensing data acquisition module collects soil moisture data through a soil moisture sensor and monitors crop growth data to synthesize real-time soil and growth data; performs time series analysis on the real-time soil and growth data to identify data trends and abnormal points and generate trend analysis results; a pest and disease analysis module, receiving the trend analysis result, analyzing the pest and disease signs in the crop image through image recognition and convolutional neural network, identifying the pest and disease type, and generating a pest and disease identification result; evaluating the probability of pest and disease occurrence according to the pest and disease identification result, and generating a pest and disease occurrence probability evaluation result; The irrigation control module adjusts the irrigation amount according to the pest occurrence probability assessment result, optimizes the water pump rate and switching time to match the current crop demand and the predicted pest risk, and generates adjusted irrigation parameters; combines the matching degree of the adjusted irrigation parameters with the real-time soil moisture data, optimizes water distribution, and obtains the optimal irrigation configuration; The fertilization decision module is based on the irrigation optimization configuration and collects real-time data of soil moisture to determine the need for chemical fertilizer adjustment, and selects the type and application amount of chemical fertilizer based on the growth requirements of crops and the chemical properties of the soil.

2. The variable fertilization control system for tobacco fields according to claim 1, characterized in that: The steps for obtaining the real-time soil and growth data are as follows: Soil moisture sensors are used to monitor soil moisture in tobacco fields, and cameras are used to capture crop growth status to obtain preliminary soil moisture data and crop growth data. Based on the preliminary soil moisture data and crop growth data, aligning and merging time stamps to form a unified data set; Based on the unified data set, the interaction between soil moisture and crop growth is analyzed, and real-time soil and growth data are generated through fusion processing.

3. The variable fertilization control system for tobacco fields according to claim 1, characterized in that: The steps for obtaining the trend analysis results are: Based on the real-time soil and growth data, statistical analysis is performed, including calculation of mean, standard deviation and extreme value, to obtain statistical characteristic analysis results of the data; Based on the statistical characteristics analysis results of the data, time series analysis is performed to calculate the comprehensive score at each time point. The calculation formula is: ； in, is the comprehensive score at time t, is the observed value at time t, and are the mean and standard deviation of the data set, respectively. is the moving average, is the standard deviation of the moving average, and is the weight coefficient; Based on the comprehensive score, the overall trend is evaluated and abnormal points are identified to obtain trend analysis results.

4. The variable fertilization control system for tobacco fields according to claim 1, characterized in that: The steps for obtaining the pest identification result are as follows: Using the trend analysis results, selecting crop images for pest and disease sign analysis to obtain selected crop images; Based on the selected crop image, deploy a convolutional neural network to perform image processing, identify pest and disease characteristics in the image, and obtain a pest and disease characteristic analysis result in the image; Based on the analysis results of the pest and disease characteristics in the image, the pest and disease types are classified, and by comparing with an established pest and disease characteristic database, the pest and disease types are identified and calibrated to obtain pest and disease identification results.

5. The variable fertilization control system for tobacco fields according to claim 1, characterized in that: The steps for obtaining the pest occurrence probability assessment result are as follows: Receiving the pest identification result, including the type of each pest and the distribution characteristics in the image, and obtaining pest distribution data; Based on the pest distribution data, the occurrence probability of various pests is calculated, and the calculation formula is: ； in, is the occurrence probability of the i-th pest, Represents the coverage area of ​​pests and diseases in the identification results, Represents the degree of impact, represents temperature sensitivity, Represents seasonality, is the adjustment coefficient, n is the total number of pests and diseases; Based on the occurrence probability, a risk assessment is conducted, and combined with environmental factors and past data, an assessment result of the occurrence probability of pests and diseases is obtained.

6. The variable fertilization control system for tobacco fields according to claim 1, characterized in that: The steps for obtaining the adjusted irrigation parameters are: Analyze the probability of each type of pests and diseases provided by the pests and diseases occurrence probability assessment results, and conduct an impact analysis on irrigation demand based on the probability, including investigation of crop growth cycles and water requirements, to determine the range of initial irrigation volume to be adjusted; Based on the preliminary irrigation volume range, calculate the pump rate using the formula: ； in, is the adjusted pump speed, is the basic pump speed, is the amount of water currently required by crops, is the standard water volume, is the pest risk adjustment factor; Based on the water pump rate, the irrigation parameters are reconfigured to obtain adjusted irrigation parameters.

7. The variable fertilization control system for tobacco fields according to claim 1, characterized in that: The steps for obtaining the optimal irrigation configuration are: Based on the adjusted irrigation parameters and soil moisture data, the matching index between the two is calculated, and the calculation formula is: ； in, is the matching index, is the ideal soil moisture at the i-th detection point, is the soil moisture at the i-th detection point, is the number of detection points; The operating parameters of the water pump and the irrigation plan are optimized according to the matching index to obtain an optimal irrigation configuration.

8. A variable fertilization control method for tobacco fields, characterized in that: The variable fertilization control system for tobacco fields according to any one of claims 1 to 7 comprises the following steps: Soil moisture data is collected through soil moisture sensors, and crop growth data is collected at the same time. Time series analysis is performed on the collected data to identify data trends and anomalies, and obtain comprehensive analysis results of soil and growth data; Based on the comprehensive analysis results of soil and growth data, image recognition and convolutional neural networks are used to analyze the signs of pests and diseases in crop images, identify the types of pests and diseases, assess the probability of pest and disease occurrence, and generate pest and disease identification and risk assessment results; Based on the results of pest and disease identification and risk assessment, adjust the irrigation amount, optimize the pump rate and switching time to match the current crop demand and predicted pest and disease risks, and generate adjusted irrigation parameters; Combine the adjusted irrigation parameters with the real-time soil moisture data to optimize water distribution and obtain the optimal irrigation configuration result; Using the irrigation optimization configuration results, the type and application amount of chemical fertilizers are selected according to the growth requirements of crops and the chemical properties of the soil.

Images