A citrus fruit cracking prevention and control method based on intelligent monitoring and regulation

Abstract

The present application relates to the technical field of citrus fruit cracking prevention and control, and proposes a citrus fruit cracking prevention and control method based on intelligent monitoring and regulation, comprising the following steps: collecting soil humidity data, fruit growth data and meteorological condition data of the monitored citrus; identifying the soil humidity data through a soil humidity analysis model to obtain a soil humidity evaluation state; identifying the fruit growth data through a fruit growth analysis model to obtain a fruit growth evaluation result; identifying the meteorological condition data through a meteorological condition risk evaluation model to obtain a meteorological condition influence degree; identifying the soil humidity evaluation state, the fruit growth evaluation result and the meteorological condition influence degree through a comprehensive risk judgment model to obtain a citrus fruit cracking risk degree of the monitored citrus; controlling the irrigation system and the fertilization device of the monitored citrus based on the citrus fruit cracking risk degree to adjust the irrigation amount, the fertilization amount and the ventilation equipment; and sending an early warning to the grower when the citrus fruit cracking risk degree exceeds a preset threshold value so as to take prevention and control measures. The present application realizes the accuracy of citrus fruit cracking risk evaluation, thereby improving the management efficiency of citrus planting and the yield and quality of the fruit.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of citrus fruit cracking prevention and control, and particularly relates to a citrus fruit cracking prevention and control method based on intelligent monitoring and regulation. BACKGROUND

[0002] Citrus fruit cracking is a common problem in citrus cultivation, which seriously affects the yield and quality of citrus. At present, the prevention and control of citrus fruit cracking mainly relies on traditional planting management experience, and lacks scientific and effective precision prevention and control means.

[0003] In the prior art, the monitoring and regulation of the growth environment of citrus are often not accurate and timely enough to meet the actual needs of citrus growth. Traditional irrigation and fertilization methods are usually based on experience and fixed schedules, and cannot be adjusted according to the real-time growth conditions and environmental changes of citrus. The real-time risk assessment effect of citrus fruit cracking is poor, which leads to unstable soil moisture and nutrient supply, and increases the risk of fruit cracking. SUMMARY

[0004] Therefore, the present application provides a citrus fruit cracking prevention and control method based on intelligent monitoring and regulation, which solves the problem of poor real-time risk assessment effect of citrus fruit cracking in the prior art.

[0005] The technical scheme of the present application is as follows: The present application provides a citrus fruit cracking prevention and control method based on intelligent monitoring and regulation, comprising the following steps:

[0006] S1, collecting original soil moisture data, original fruit growth data and original weather condition data of the monitored citrus through a sensor network, preprocessing the original soil moisture data, original fruit growth data and original weather condition data to obtain soil moisture data, fruit growth data and weather condition data;

[0007] S2, constructing a soil moisture analysis model based on a logistic regression model, identifying the soil moisture data through the soil moisture analysis model to obtain a soil moisture evaluation state;

[0008] S3, constructing a fruit growth analysis model according to a deep neural network, identifying the fruit growth data through the fruit growth analysis model to obtain a fruit growth evaluation result;

[0009] S4, constructing a weather condition risk assessment model using a decision tree model, identifying the weather condition data through the weather condition risk assessment model to obtain a weather condition influence degree;

[0010] S5, constructing a comprehensive risk judgment model by a random forest model, identifying the soil humidity evaluation state, the fruit growth evaluation result and the meteorological condition influence degree through the comprehensive risk judgment model, and obtaining the citrus fruit cracking risk degree of the monitored citrus;

[0011] S6, controlling the irrigation system and the fertilization device of the monitored citrus based on the citrus fruit cracking risk degree, adjusting the irrigation amount, the fertilization amount and the ventilation equipment;

[0012] S7, when the citrus fruit cracking risk degree exceeds a preset threshold, sending a warning to a grower to take prevention and control measures.

[0013] In the above technical scheme, preferably, step S1 comprises:

[0014] The original soil humidity data is collected by a soil humidity sensor, and the original soil humidity data is filtered and normalized to obtain soil humidity data. The original fruit growth data is collected by a fruit growth monitor, and the original fruit growth data is smoothed and standardized to obtain fruit growth data. The original meteorological condition data is collected by a meteorological monitoring device, and the original meteorological condition data is missing value filled and outlier detected to obtain meteorological condition data.

[0015] In the above technical scheme, preferably, step S2 comprises:

[0016] The historical soil humidity data and the citrus growth demand data are obtained, and a soil humidity analysis model is constructed based on a logistic regression model, the historical soil humidity data and the citrus growth demand data.

[0017] The soil moisture content and the soil temperature are selected as characteristic variables and the regression coefficient is determined, the historical soil humidity data and the citrus growth demand data are used to train the soil humidity analysis model, and the regression coefficient is adjusted to minimize the prediction error.

[0018] The trained soil humidity analysis model is verified by a cross-validation method to evaluate the performance of the model on different data sets, and the soil humidity analysis model is obtained.

[0019] The soil humidity data is input into the soil humidity analysis model to obtain the soil humidity evaluation state.

[0020] In the above technical scheme, preferably, step S3 comprises:

[0021] An initial fruit growth analysis model is constructed based on a deep neural network, historical fruit growth data and a growth rule of the citrus are obtained, the initial fruit growth analysis model is continuously trained based on the historical fruit growth data and the growth rule of the citrus, and a fruit growth analysis model is obtained.

[0022] Based on the above technical solutions, preferably, the calculation formula of the fruit growth analysis model is:

[0023] G=σ(W3·σ(W2·σ(W1·X+b1)+b2)+b3);

[0024]

[0025] wherein G is a fruit growth evaluation result, X is fruit growth data, respectively, W1, W2, and W3 are a weight matrix of fruit size, a weight matrix of fruit weight, and a weight matrix of fruit growth speed, respectively, σ(·) is an activation function, b1, b2, and b3 are a bias constant of fruit size, a bias constant of fruit weight, and a bias constant of fruit growth speed, respectively;

[0026] The fruit growth data is input into the fruit growth analysis model, and a fruit growth evaluation result is obtained.

[0027] Based on the above technical solutions, preferably, step S4 comprises:

[0028] A citrus meteorological factor affecting the citrus is obtained, meteorological characteristic variables of the citrus meteorological factor are extracted, a decision tree model is constructed based on the meteorological characteristic variables, historical meteorological data are obtained, the decision tree model is iteratively trained based on the historical meteorological data until an iteration stopping condition is reached, a trained decision tree model is obtained, the trained decision tree model is verified and evaluated through cross-validation, and a meteorological condition risk evaluation model is obtained.

[0029] The meteorological condition data are identified through the meteorological condition risk evaluation model, and a meteorological condition influence degree is obtained.

[0030] Based on the above technical solutions, preferably, step S5 comprises:

[0031] A random forest model is constructed taking soil humidity, fruit growth, and meteorological conditions as characteristic variables, the random forest model is trained based on multi-source historical data, the multi-source historical data include a historical soil humidity characteristic vector, a historical fruit growth characteristic vector, and a historical meteorological condition characteristic vector, the trained random forest model is verified and evaluated through cross-validation, and a comprehensive risk judgment model is obtained.

[0032] On the basis of the above technical scheme, preferably, the comprehensive risk judgment model is calculated by the following formula:

[0033]

[0034] Wherein, Q is the risk degree of citrus fruit cracking, U is the number of decision trees in the random forest model, θ u is the fitting parameter of the u-th decision tree, P(y=k|x, θ u ) is the prediction probability of the u-th decision tree for the sample x belonging to the k-th risk degree.

[0035] The soil moisture evaluation state, the fruit growth evaluation result and the meteorological condition influence degree are identified by the comprehensive risk judgment model, the risk degree of citrus fruit cracking of the monitored citrus is obtained, and the corresponding risk level is output.

[0036] On the basis of the above technical scheme, preferably, step S6 comprises:

[0037] According to the risk degree of citrus fruit cracking, the irrigation amount of the irrigation system is set, when the risk degree of citrus fruit cracking is higher than the preset high risk threshold, the irrigation amount is increased, and when the risk degree of citrus fruit cracking is lower than the preset low risk threshold, the irrigation amount is reduced.

[0038] The fertilization amount of the fertilization device is adjusted, according to the risk evaluation result, when the risk degree of citrus fruit cracking is higher than the preset high risk threshold, the fertilization amount is increased, and when the risk degree of citrus fruit cracking is lower than the preset low risk threshold, the fertilization amount is reduced.

[0039] The opening and closing of the ventilation equipment is controlled, when the risk degree of citrus fruit cracking is higher than the preset high risk threshold, the ventilation equipment is opened to reduce the internal humidity of the fruit, and in the low risk case, the ventilation equipment is closed to maintain the growth environment.

[0040] On the basis of the above technical scheme, preferably, step S7 comprises:

[0041] When the risk degree of citrus fruit cracking is monitored to exceed the preset threshold, the warning information is sent to the planting personnel through short message, email or APP push, reminding the planting personnel to take corresponding prevention and control measures, such as increasing irrigation frequency, adjusting fertilization scheme and opening ventilation equipment.

[0042] The citrus fruit cracking prevention and control method based on intelligent monitoring and regulation of the present application has the following beneficial effects compared with the prior art:

[0043] (1) By collecting soil moisture, fruit growth and weather condition data in real time, using soil moisture analysis model, fruit growth analysis model and weather condition risk assessment model for identification respectively, and then through the comprehensive risk judgment model for risk assessment, the accuracy of citrus fruit cracking risk assessment is improved, and according to the evaluation results, irrigation, fertilization and ventilation equipment are automatically adjusted to optimize the growth environment, and at the same time, when the risk exceeds the threshold, early warning information is sent to the grower in time, so as to improve the management efficiency of citrus planting and the yield and quality of fruit;

[0044] (2) By integrating the prediction results of random forest model and multiple decision trees, combining dynamic weight adjustment and multi-class risk prediction, comprehensive and accurate assessment of citrus fruit cracking risk is realized, not only the accuracy of risk assessment is improved, but also more detailed risk analysis and timely decision support is provided for the grower, so as to effectively optimize the citrus planting management strategy and reduce the fruit cracking risk;

[0045] (3) By constructing a fruit growth analysis model based on deep neural network, combining historical fruit growth data and training of citrus growth law, accurate assessment of fruit growth state is realized, the accuracy of fruit growth assessment is improved, and abnormal growth conditions can be identified in time. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0047] Figure 1 A flow chart of a citrus fruit cracking prevention and control method based on intelligent monitoring and control. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the scope of protection of the present application.

[0049] Please refer to Figure 1 The present application provides a citrus fruit cracking prevention and control method based on intelligent monitoring and control, comprising the following steps:

[0050] S1, collecting original soil humidity data, original fruit growth data and original weather condition data of the monitored citrus through a sensor network, preprocessing the original soil humidity data, the original fruit growth data and the original weather condition data to obtain soil humidity data, fruit growth data and weather condition data;

[0051] S2, constructing a soil humidity analysis model based on a logistic regression model, identifying the soil humidity data through the soil humidity analysis model to obtain a soil humidity evaluation state;

[0052] S3, constructing a fruit growth analysis model according to a deep neural network, identifying the fruit growth data through the fruit growth analysis model to obtain a fruit growth evaluation result;

[0053] S4, constructing a weather condition risk evaluation model using a decision tree model, identifying the weather condition data through the weather condition risk evaluation model to obtain a weather condition influence degree;

[0054] S5, constructing a comprehensive risk judgment model through a random forest model, identifying the soil humidity evaluation state, the fruit growth evaluation result and the weather condition influence degree through the comprehensive risk judgment model to obtain a citrus fruit cracking risk degree of the monitored citrus;

[0055] S6, controlling an irrigation system and a fertilization device of the monitored citrus based on the citrus fruit cracking risk degree to adjust irrigation amount, fertilization amount and ventilation equipment;

[0056] S7, when the citrus fruit cracking risk degree exceeds a preset threshold, sending a warning to a grower to take prevention and control measures.

[0057] Specifically, the embodiment collects soil humidity, fruit growth and weather condition data in real time, identifies them respectively using a soil humidity analysis model, a fruit growth analysis model and a weather condition risk evaluation model, then performs risk evaluation through a comprehensive risk judgment model, improves the accuracy of citrus fruit cracking risk evaluation, automatically adjusts irrigation, fertilization and ventilation equipment according to the evaluation result, optimizes the growth environment, and sends warning information to the grower in time when the risk exceeds the threshold, thereby improving the management efficiency of citrus planting and the yield and quality of fruits.

[0058] Step S1 includes:

[0059] The original soil moisture data is collected by a soil moisture sensor, and the original soil moisture data is filtered and normalized to obtain soil moisture data. The original fruit growth data is collected by a fruit growth monitor, and the original fruit growth data is smoothed and standardized to obtain fruit growth data. The original weather condition data is collected by a weather monitoring device, and the original weather condition data is missing value filled and outlier detected to obtain weather condition data.

[0060] In a specific embodiment, the original soil moisture data is collected periodically by a soil moisture sensor, and a filtering algorithm is used to remove noise in the data, followed by normalization to standardize the data to the range of 0 to 1.

[0061] The original fruit growth data is collected periodically by a fruit growth monitor, and a smoothing method is used to reduce data fluctuations, and a standardization process is performed to make data at different time points comparable.

[0062] The original weather condition data is collected by a weather monitoring device, and missing value filling (such as using mean filling method) and outlier detection (such as using Z-score method) are performed to ensure the integrity and accuracy of the weather data, and finally reliable weather condition data is obtained.

[0063] Specifically, the embodiment improves the quality of soil moisture, fruit growth and weather condition data by filtering, normalizing, smoothing and standardizing the original data, integrates multi-source data of soil moisture, fruit growth and weather condition, forms a comprehensive monitoring system, and better reflects the actual growth conditions. By regularly collecting and real-time processing data, the response ability to changes in citrus growing environment is enhanced, so that the planting personnel can timely adjust the management measures to reduce the risk of fruit cracking.

[0064] Step S2 includes:

[0065] Obtain historical soil moisture data and citrus growth demand data, and construct a soil moisture analysis model based on a logistic regression model, the historical soil moisture data and the citrus growth demand data;

[0066] Select soil moisture content and soil temperature as feature variables and determine regression coefficients, train the soil moisture analysis model using historical soil moisture data and citrus growth demand data, and adjust the regression coefficients to minimize the prediction error;

[0067] Use cross-validation method to verify the trained soil moisture analysis model to evaluate the performance of the model on different data sets, and obtain the soil moisture analysis model;

[0068] The calculation formula of the soil moisture analysis model is:

[0069]

[0070] where H is the soil moisture assessment state, W is the soil water content, T is the soil temperature, a, b, c are the regression coefficients for soil water content, soil temperature, and bias constant, respectively.

[0071] The soil moisture data is input into the soil moisture analysis model to obtain the soil moisture assessment state H.

[0072] In a specific embodiment, first, historical soil moisture data and citrus growth requirement data are collected. These data can be obtained through long-term monitoring and recording, covering soil moisture and citrus growth requirements at different growth stages.

[0073] Based on the collected historical soil moisture data and citrus growth requirement data, a soil moisture analysis model is constructed using a logistic regression model. Soil water content and soil temperature are selected as feature variables, and the corresponding regression coefficients are determined.

[0074] The soil moisture analysis model is trained using historical soil moisture data and citrus growth requirement data, and the regression coefficients are adjusted to minimize prediction error. The cross-validation method is used to verify the trained model to evaluate the model's performance on different data sets.

[0075] The real-time collected soil moisture data is input into the trained soil moisture analysis model to obtain the soil moisture assessment state H, reflecting the current soil moisture impact on citrus growth.

[0076] Specifically, by using historical data for training and verification, the logistic regression model can more accurately reflect the relationship between soil moisture and citrus growth requirements. By selecting soil water content and soil temperature as feature variables, the model can more effectively capture key factors affecting citrus growth. By inputting real-time soil moisture data into the trained model, the soil moisture assessment state can be quickly obtained.

[0077] The soil moisture analysis model is constructed based on a logistic regression model, which is a machine learning algorithm commonly used for binary classification problems. In this embodiment, the soil moisture assessment state H is considered as a binary classification output (high / low humidity).

[0078] Soil water content W and soil temperature T are selected as feature variables. These two factors are key environmental parameters that affect soil moisture and can better reflect the soil moisture state.

[0079] The three regression coefficients a, b, and c are determined by training a logistic regression model using historical soil moisture data and citrus growth requirement data. These coefficients reflect the influence of each feature variable on the soil moisture assessment state.

[0080] The trained soil moisture analysis model is validated and evaluated using a cross-validation method. This ensures that the model performs stably and reliably on different datasets.

[0081] Step S3 includes:

[0082] An initial fruit growth analysis model is constructed based on a deep neural network. Historical fruit growth data and citrus growth patterns are obtained, and the initial fruit growth analysis model is further trained based on the historical fruit growth data and the citrus growth patterns to obtain a fruit growth analysis model.

[0083] The calculation formula of the fruit growth analysis model is:

[0084] G = σ(W3·σ(W2·σ(W1·X+b1)+b2)+b3);

[0085]

[0086] where G is the fruit growth assessment result, X is the fruit growth data, respectively, W1, W2, and W3 are the weight matrix of fruit size, the weight matrix of fruit weight, and the weight matrix of fruit growth speed, respectively, σ(·) is the activation function, b1, b2, and b3 are the bias constant of fruit size, the bias constant of fruit weight, and the bias constant of fruit growth speed, respectively.

[0087] The fruit growth data is input into the fruit growth analysis model to obtain the fruit growth assessment result.

[0088] In a specific embodiment, historical fruit growth data and citrus growth patterns are collected, including information such as fruit size, weight, and growth speed. These data can be obtained through long-term monitoring and recording, covering fruit growth at different growth stages.

[0089] An initial fruit growth analysis model is constructed based on a deep neural network, with fruit size, fruit weight, and fruit growth speed as input features, and corresponding weight matrices and bias constants are set.

[0090] The initial fruit growth analysis model is trained using historical fruit growth data and citrus growth patterns, and the weight matrices and bias constants are adjusted to minimize prediction errors. The trained model is validated using a cross-validation method to evaluate the model's performance on different datasets.

[0091] The real-time collected fruit growth data is input into the trained fruit growth analysis model to obtain a fruit growth evaluation result G reflecting the current fruit growth state.

[0092] Specifically, by using a deep neural network, the model can capture the complex nonlinear relationship in the fruit growth process, and the fruit size, weight and growth rate are selected as input features, so that the model can comprehensively reflect the growth state of the fruit.

[0093] The fruit growth analysis model is constructed based on a deep neural network, which can automatically learn the complex nonlinear relationship between features, thereby better capturing the rules of fruit growth.

[0094] The model considers three characteristic variables of fruit size, fruit weight and fruit growth rate, which can comprehensively reflect the growth state of the fruit.

[0095] By training and adjusting the weight matrices W1, W2, W3 and the bias constants b1, b2, b3, the model can automatically learn the relative importance of different features to fruit growth, thereby making more accurate predictions.

[0096] The introduction of an activation function σ(·), such as a sigmoid function or a ReLU function, in the model output can introduce a nonlinear transformation and enhance the expression ability of the model.

[0097] Step S4 includes:

[0098] Obtain citrus meteorological factors affecting citrus, extract meteorological characteristic variables of the citrus meteorological factors, construct a decision tree model based on the meteorological characteristic variables, obtain historical meteorological data, and iteratively train the decision tree model through the historical meteorological data until an iteration stopping condition is reached to obtain a trained decision tree model, validate and evaluate the trained decision tree model through cross-validation to obtain a meteorological condition risk assessment model;

[0099] The calculation formula of the meteorological condition risk assessment model is:

[0100]

[0101] wherein R is the meteorological condition influence degree, w i is the weight coefficient of the i-th node in the decision tree, X i is the characteristic value of the i-th citrus meteorological factor, n is the number of nodes of the decision tree, and exp(·) is an exponential function.

[0102] The meteorological condition risk assessment model is used to identify the meteorological condition data to obtain the meteorological condition influence degree.

[0103] In a specific embodiment, n is 5, and meteorological factors affecting citrus growth are collected, including temperature, humidity, precipitation, wind speed, and light. These data can be collected regularly by meteorological monitoring equipment to ensure the timeliness and accuracy of the data.

[0104] Meteorological characteristic variables such as soil temperature, air humidity, and precipitation are extracted from the collected meteorological factors as input features of the decision tree model.

[0105] A decision tree model is constructed based on the extracted meteorological characteristic variables, and the model is trained using historical meteorological data to determine the weight coefficients and feature values of each node until a preset iteration stopping condition is reached.

[0106] The trained decision tree model is verified and evaluated by cross-validation method to ensure the accuracy and reliability of the model, and finally a meteorological condition risk assessment model is obtained.

[0107] The trained meteorological condition risk assessment model is used to identify meteorological condition data, and the meteorological condition influence degree R is obtained, reflecting the influence of meteorological factors on citrus growth.

[0108] Specifically, the present embodiment forms a comprehensive meteorological monitoring system by collecting various meteorological factors, can timely identify potential meteorological risks, analyzes the meteorological conditions using the decision tree model, can clearly show the influence of different meteorological factors on citrus growth, and through the analysis of real-time meteorological data, can quickly evaluate the influence of meteorological conditions on citrus growth, providing timely risk warning for planting personnel, and promoting rapid response.

[0109] The meteorological condition risk assessment model is constructed based on the decision tree model, and by introducing the exponential function exp, the nonlinear influence of meteorological factors on citrus fruit cracking risk can be better reflected. The exponential function can amplify the importance of meteorological factors and improve the model's ability to identify extreme weather conditions. The decision tree model can automatically learn the complex relationship between meteorological factors and citrus fruit cracking risk without manual rule setting, improving the model's adaptability and generalization ability.

[0110] Step S5 includes:

[0111] Soil humidity, fruit growth, and meteorological conditions are used as characteristic variables to construct a random forest model, and the random forest model is trained based on multi-source historical data including historical soil humidity feature vectors, historical fruit growth feature vectors, and historical meteorological condition feature vectors. The trained random forest model is verified and evaluated by cross-validation to obtain a comprehensive risk judgment model.

[0112] The calculation formula of the comprehensive risk judgment model is:

[0113]

[0114] where Q is the citrus fruit cracking risk level, U is the number of decision trees in the random forest model, θ u is the fitting parameter of the u-th decision tree, P(y=k|x, θ u ) is the prediction probability of the u-th decision tree that the sample x belongs to the k-th risk level.

[0115] The soil moisture evaluation state, the fruit growth evaluation result, and the weather condition influence degree are identified by the comprehensive risk judgment model to obtain the citrus fruit cracking risk level of the monitored citrus, and the corresponding risk level is output.

[0116] In a specific embodiment, soil moisture, fruit growth evaluation result, and weather condition influence degree are selected as feature variables to construct a random forest model. These feature variables can comprehensively reflect the environment and state of citrus growth.

[0117] Based on multi-source historical data, including historical soil moisture feature vectors, historical fruit growth feature vectors, and historical weather condition feature vectors, a random forest model is constructed. Each decision tree of the random forest model uses different feature subsets and sample subsets during training to enhance the generalization ability of the model.

[0118] The random forest model is trained using historical data, and the model parameters are adjusted to improve prediction accuracy. The trained random forest model is verified and evaluated by cross-validation method to ensure the stability and reliability of the model.

[0119] The soil moisture evaluation state, the fruit growth evaluation result, and the weather condition influence degree are identified by the trained random forest model to obtain the citrus fruit cracking risk level Q, and the corresponding risk level is output.

[0120] The risk level Q is usually a continuous value representing the risk level of citrus fruit cracking. This value can be a number between 0 and 1, where 0 represents no risk and 1 represents extremely high risk.

[0121] According to the obtained risk level Q, multiple risk levels can be set to facilitate the understanding and adoption of corresponding measures by planting personnel. Common risk level divisions can include:

[0122] Low risk (Q<0.3): indicates a low risk of fruit cracking, and the current management measures basically meet the needs of citrus growth.

[0123] Medium risk (0.3≤Q<0.7): indicates a medium risk of fruit cracking, and it is recommended that planting personnel pay attention to soil moisture and weather changes and adjust management measures appropriately.

[0124] High risk (Q ≥ 0.7): indicates a higher risk of fruit cracking, and immediate preventive measures should be taken, such as increasing irrigation, adjusting fertilization, and turning on ventilation equipment, etc.

[0125] Specifically, by integrating multi-source data of soil moisture, fruit growth, and meteorological conditions, the random forest model can comprehensively assess the risk of citrus fruit cracking. By constructing multiple decision trees and voting, the random forest model can effectively reduce overfitting. By inputting real-time soil moisture, fruit growth, and meteorological condition data, the risk level of citrus fruit cracking can be quickly assessed, providing timely risk warnings for growers.

[0126] The comprehensive risk judgment model is based on the random forest model, which integrates the prediction results of multiple decision trees. By utilizing the independence and diversity of each tree, the stability and accuracy of the model are enhanced. This integrated method can effectively reduce the bias and variance of a single model.

[0127] Different fitting parameters θ are assigned to each decision tree u According to its performance on the training set, the weight is dynamically adjusted, so that the better-performing decision trees contribute more to the final risk assessment, thereby improving the predictive ability of the overall model.

[0128] By synthesizing the prediction probabilities of the sample x belonging to different risk categories k, the risk level of citrus fruit cracking can be more comprehensively assessed, providing more detailed risk analysis.

[0129] Step S6 includes:

[0130] According to the risk level of citrus fruit cracking, the irrigation amount of the irrigation system is set. When the risk level of citrus fruit cracking is higher than the preset high risk threshold, the irrigation amount is increased; when the risk level of citrus fruit cracking is lower than the preset low risk threshold, the irrigation amount is reduced.

[0131] Adjust the fertilization amount of the fertilization device. According to the risk assessment result, when the risk level of citrus fruit cracking is higher than the preset high risk threshold, increase the fertilization amount; when the risk level of citrus fruit cracking is lower than the preset low risk threshold, reduce the fertilization amount.

[0132] Control the opening and closing of the ventilation equipment. When the risk level of citrus fruit cracking is higher than the preset high risk threshold, turn on the ventilation equipment to reduce the internal humidity of the fruit; in the low risk case, turn off the ventilation equipment to maintain the growing environment.

[0133] In a specific embodiment, according to the evaluation result of the risk level of citrus fruit cracking, the irrigation amount of the irrigation system is set. When the monitored risk level is higher than the preset high risk threshold, increase the irrigation amount to ensure that the soil maintains appropriate humidity; when the risk level is lower than the preset low risk threshold, reduce the irrigation amount to avoid over-wetting and cause fruit cracking.

[0134] According to the risk assessment results, the fertilization amount of the fertilization device is adjusted. When the citrus fruit cracking risk degree is higher than the preset high risk threshold, the fertilization amount is increased to promote the healthy growth of the fruit; when the risk degree is lower than the preset low risk threshold, the fertilization amount is reduced to avoid nutrient excess.

[0135] According to the citrus fruit cracking risk degree, the opening and closing of the ventilation equipment are controlled. When the risk degree is higher than the preset high risk threshold, the ventilation equipment is turned on to reduce the internal humidity of the fruit and reduce the cracking risk; in the low risk case, the ventilation equipment is turned off to maintain a suitable growth environment.

[0136] Through the monitoring system, the effects of irrigation, fertilization and ventilation are tracked in real time, and the management strategy is adjusted in time to ensure the best state of the citrus growth environment.

[0137] Specifically, the embodiment realizes precise management of the citrus growth environment by dynamically adjusting irrigation, fertilization and ventilation measures according to the citrus fruit cracking risk degree, reduces the cracking risk, and improves the growth quality of the fruit. According to the real-time risk assessment results, the management measures are adjusted, which can effectively avoid waste of resources, such as reducing unnecessary use of water and fertilizer. Through the real-time monitoring and feedback mechanism, the planting personnel can quickly respond to environmental changes.

[0138] Step S7 includes:

[0139] When it is monitored that the citrus fruit cracking risk degree exceeds the preset threshold, an early warning information is sent to the planting personnel through SMS, email or APP push to remind the planting personnel to take corresponding prevention and control measures, such as increasing irrigation frequency, adjusting fertilization scheme, and turning on the ventilation equipment.

[0140] In a specific embodiment, according to historical data and expert experience, the preset threshold of citrus fruit cracking risk is set, including the high risk threshold and the low risk threshold. These thresholds will serve as the basis for risk assessment to take appropriate prevention and control measures in time.

[0141] When the monitored citrus fruit cracking risk degree exceeds the preset high risk threshold, the early warning mechanism will be triggered automatically. The early warning information can be sent to the planting personnel through SMS, email or mobile application (mobile APP) push.

[0142] The early warning information should include the current risk assessment results, the recommended prevention and control measures (such as increasing irrigation frequency, adjusting fertilization scheme, turning on the ventilation equipment, etc.), and the relevant weather forecast and environmental change information to help the planting personnel make timely decisions.

[0143] After receiving the early warning information, the planting personnel can quickly take corresponding prevention and control measures to ensure that the growth environment of the citrus is adjusted in time to reduce the cracking risk.

[0144] Specifically, the embodiment can discover potential crack risks in time by setting risk threshold and establishing monitoring mechanism, ensure that the planting personnel can take preventive measures before the problem occurs, reduce the loss, realize real-time information transmission to the planting personnel by using modern communication technology (such as short message, email and APP push), so that the planting personnel can quickly obtain important information, the suggestion prevention and control measures contained in the early warning information provide scientific decision support for the planting personnel, help them make reasonable management adjustment when facing risks, through the establishment of a comprehensive monitoring and early warning system, the planting personnel can more effectively manage risks, optimize resource allocation, and improve overall production efficiency and economic benefits.

[0145] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for preventing and controlling citrus fruit cracking based on intelligent monitoring and regulation, characterized in that, The method comprises the following steps: S1, collecting original soil moisture data, original fruit growth data and original weather condition data of the monitored citrus through a sensor network, preprocessing the original soil moisture data, the original fruit growth data and the original weather condition data to obtain soil moisture data, fruit growth data and weather condition data; S2, constructing a soil moisture analysis model based on a logistic regression model, identifying the soil moisture data through the soil moisture analysis model to obtain a soil moisture evaluation state; The calculation formula of the soil moisture analysis model is: ； wherein, H is a soil moisture assessment state, W is a soil moisture content, T is a soil temperature, a , b , c are a regression coefficient of the soil moisture content, a regression coefficient of the soil temperature, a bias constant, respectively. S3, constructing a fruit growth analysis model according to a deep neural network, identifying the fruit growth data through the fruit growth analysis model to obtain a fruit growth evaluation result; The calculation formula of the fruit growth analysis model is: ； ； wherein, G is a result of fruit growth evaluation, X is fruit growth data, , , are fruit size, fruit weight, fruit growth speed, respectively, W 1, W 2, W 3 are weight matrix of fruit size, weight matrix of fruit weight, weight matrix of fruit growth speed, respectively, is an activation function, b 1, b 2, b 3 are bias constant of fruit size, bias constant of fruit weight, bias constant of fruit growth speed, respectively; S4, constructing a weather condition risk evaluation model using a decision tree model, identifying the weather condition data through the weather condition risk evaluation model to obtain a weather condition influence degree; The calculation formula of the weather condition risk evaluation model is: ； wherein, R is the degree of influence of the weather condition, is the weight coefficient of the i th node in the decision tree, is the feature value of the i th citrus weather factor, n is the number of nodes of the decision tree, is an exponential function; S5, constructing a comprehensive risk judgment model through a random forest model, identifying the soil moisture evaluation state, the fruit growth evaluation result and the weather condition influence degree through the comprehensive risk judgment model to obtain a citrus fruit cracking risk degree of the monitored citrus; The calculation formula of the comprehensive risk judgment model is: ； wherein, Q is the degree of citrus fruit cracking risk, U is the number of decision trees in the random forest model, is the fitting parameter of the u th decision tree, is the prediction probability of the u th decision tree that the sample x belongs to the k th degree of risk; and outputting a risk level corresponding to the citrus fruit cracking risk degree; S6, controlling an irrigation system and a fertilization device of the monitored citrus based on the citrus fruit cracking risk degree to adjust irrigation amount, fertilization amount and ventilation equipment; S7, when the citrus fruit cracking risk degree exceeds a preset threshold, sending a warning to a grower to take prevention and control measures.

2. The method for preventing and controlling citrus fruit cracking based on intelligent monitoring and control as described in claim 1, characterized in that, Step S1 comprises: Collecting original soil moisture data through a soil moisture sensor and performing filtering processing and normalization processing on the original soil moisture data to obtain soil moisture data, collecting original fruit growth data through a fruit growth monitor and performing smoothing processing and standardization processing on the original fruit growth data to obtain fruit growth data, and collecting original weather condition data through a weather monitoring device and performing missing value filling and outlier detection on the original weather condition data to obtain weather condition data.

3. The method of claim 1, wherein the method comprises: monitoring the citrus fruit with the monitoring device; and controlling the citrus fruit with the control device. Step S2 comprises: Obtaining historical soil moisture data and citrus growth demand data, constructing a soil moisture analysis model based on a logistic regression model, the historical soil moisture data and the citrus growth demand data; Selecting soil moisture content and soil temperature as characteristic variables and determining regression coefficients, training the soil moisture analysis model using historical soil moisture data and citrus growth demand data, and adjusting the regression coefficients to minimize the prediction error; Verifying the trained soil moisture analysis model using a cross-validation method to evaluate the performance of the model on different data sets to obtain the soil moisture analysis model; Inputting the soil moisture data into the soil moisture analysis model to obtain a soil moisture evaluation state.

4. The method for preventing and controlling citrus fruit cracking based on intelligent monitoring and control as described in claim 3, characterized in that, Step S3 comprises: The initial fruit growth analysis model is constructed based on a deep neural network, historical fruit growth data and a growth rule of the citrus fruit are obtained, the initial fruit growth analysis model is continuously trained based on the historical fruit growth data and the growth rule of the citrus fruit, and a fruit growth analysis model is obtained.

5. The method of claim 1, wherein the method comprises: monitoring the citrus fruit with the smart monitoring device; and controlling the citrus fruit with the smart control device. Step S4 comprises: Obtaining a citrus meteorological factor affecting the citrus fruit, extracting a meteorological feature variable of the citrus meteorological factor, constructing a decision tree model based on the meteorological feature variable, obtaining historical meteorological data, iteratively training the decision tree model based on the historical meteorological data until an iteration stopping condition is reached, obtaining a trained decision tree model, verifying and evaluating the trained decision tree model through cross-validation, and obtaining a meteorological condition risk assessment model; The meteorological condition data is identified through the meteorological condition risk assessment model, and a meteorological condition influence degree is obtained.

6. The method for preventing and controlling citrus fruit cracking based on intelligent monitoring and control as described in claim 5, characterized in that, Step S5 comprises: A random forest model is constructed with soil humidity, fruit growth and meteorological conditions as feature variables, the random forest model is trained based on multi-source historical data, the multi-source historical data includes historical soil humidity feature vectors, historical fruit growth feature vectors and historical meteorological condition feature vectors, the trained random forest model is verified and evaluated through cross-validation, and a comprehensive risk judgment model is obtained.

7. The method of claim 1, wherein the method comprises: monitoring the citrus fruit with the smart monitoring device; and controlling the citrus fruit with the smart control device. Step S6 comprises: According to the citrus fruit cracking risk degree, the irrigation amount of the irrigation system is set, when the citrus fruit cracking risk degree is higher than a preset high risk threshold, the irrigation amount is increased; when the citrus fruit cracking risk degree is lower than a preset low risk threshold, the irrigation amount is reduced; The fertilization amount of the fertilization device is adjusted, according to the risk assessment result, when the citrus fruit cracking risk degree is higher than a preset high risk threshold, the fertilization amount is increased; when the citrus fruit cracking risk degree is lower than a preset low risk threshold, the fertilization amount is reduced; The opening and closing of the ventilation equipment is controlled, when the citrus fruit cracking risk degree is higher than a preset high risk threshold, the ventilation equipment is opened to reduce the internal humidity of the fruit; in the low risk condition, the ventilation equipment is closed to maintain the growth environment.

8. The method of claim 1, wherein the method comprises: monitoring the citrus fruit with the smart monitoring device; and controlling the citrus fruit with the smart control device. Step S7 comprises: When it is monitored that the citrus fruit cracking risk degree exceeds a preset threshold, warning information is sent to the grower through short message, email or APP push to remind the grower to take corresponding prevention and control measures, such as increasing irrigation frequency, adjusting fertilization scheme and opening ventilation equipment.

Images