Wheat drip irrigation water nitrogen intelligent monitoring system based on unmanned aerial vehicle multispectral inversion

Abstract

The invention discloses a drip irrigation wheat water nitrogen intelligent monitoring system based on unmanned aerial vehicle multispectral inversion, and the system employs a multispectral sensor carried by an unmanned aerial vehicle to monitor the moisture and nitrogen content of wheat in real time, and employs a dynamic spectrum selection module to automatically select an optimal spectral band for data collection. Therefore, the most accurate crop state information can be obtained. The spectrum frequency conversion module and the adaptive spectrum amplitude modulation module in the system adjust the frequency and amplitude of spectrum data in real time, and further optimize the data quality and accuracy. Based on the high-quality data, a data analysis and decision support module generates an accurate intelligent irrigation and fertilization strategy, and water and nitrogen supply is automatically adjusted to meet the actual requirements of wheat in different growth stages, so that the growth conditions of crops are optimized, the growth quality and yield of wheat are remarkably improved, and the economic benefit is increased. Agricultural production is promoted to develop towards a more intelligent and precise direction, resource waste is reduced, and economic benefits are improved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent monitoring systems, and in particular, relates to an intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles. Background Art

[0002] In agriculture, especially for water and nitrogen management of wheat crops, spectral data can be analyzed to identify crop health, including early signs of pests and diseases. Specific spectral bands are very sensitive to changes in plant diseases and pests, which helps to detect problems in a timely manner. In addition, by analyzing the spectral reflectance of crops, their moisture content can be estimated. Insufficient or excessive moisture will affect the spectral characteristics, so spectral data can be used to monitor the moisture status of soil and crops in real time. Finally, spectral data can be used to estimate the nitrogen content of crops, because nitrogen has a significant effect on the spectral reflectance characteristics of plants. This helps to optimize fertilization strategies and ensure that crops get the nutrients they need.

[0003] Traditional spectral acquisition methods usually rely on fixed spectral bands, which have significant shortcomings when facing changes in crop growth stages and environmental conditions. Since wheat at different growth stages responds differently to spectral bands, fixed-band acquisition may not capture the true state of the crop, thus affecting the accuracy of monitoring data. For example, certain spectral bands may be less sensitive to moisture and nitrogen content at a certain growth stage, but relatively more sensitive at another stage. Therefore, fixed-band acquisition methods may result in incomplete or unrepresentative data, which in turn affects the overall monitoring effect.

[0004] In addition, many traditional systems rely on static models or a single data source when generating irrigation and fertilization strategies, which limits the adaptability and accuracy of the strategies. Static models fail to consider the changing needs of crops at different growth stages and environmental conditions, and do not fully utilize high-quality multispectral data for comprehensive analysis. Such strategies may not accurately reflect the actual needs of crops, resulting in improper use of resources, such as excessive fertilization or irrigation, affecting crop health and yield.

[0005] In view of this, the present invention is proposed. Summary of the invention

[0006] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide an intelligent water and nitrogen monitoring system for drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles, which solves the problems raised in the above-mentioned background technology.

[0007] In order to solve the above technical problems, the basic concept of the technical solution adopted by the present invention is:

[0008] An intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multispectral inversion of unmanned aerial vehicles, comprising: an unmanned aerial vehicle platform, where a sensor with multispectral imaging capability is carried;

[0009] Dynamic spectrum selection module, which is used to automatically select the most suitable spectral band for data collection according to the growth stage of crops and environmental conditions;

[0010] A spectrum frequency conversion module, which is used to convert the collected spectrum data into spectrum frequency;

[0011] An adaptive spectral amplitude modulation module, which is used to dynamically adjust the amplitude of the spectral signal during the spectral data acquisition process and automatically adjust the spectral amplitude according to real-time data feedback;

[0012] A nonlinear spectral response model module, which is used to establish a fully connected neural network model between spectral signals and wheat moisture and nitrogen content and optimize the parameters of the fully connected neural network model in real time;

[0013] Data analysis and decision support module, which is used to analyze the data processed by spectral frequency conversion and amplitude modulation, and generate intelligent irrigation and fertilization strategies based on prediction;

[0014] The control unit is used to coordinate the operation of the spectrum frequency conversion device, the adaptive spectrum amplitude modulation system and the dynamic spectrum selection module.

[0015] Optionally, the data analysis and decision support module includes a data fusion unit and a user interface module. The data fusion unit is used to comprehensively analyze the spectral data and other related data and generate optimized irrigation and fertilization plans based on the analysis results. The user interface module is used to display real-time monitoring data, analysis results and optimization suggestions, and support users in making decisions.

[0016] Optionally, the steps for automatically selecting the most suitable spectral band for data collection according to the crop growth stage and environmental conditions are:

[0017] Collect crop growth stage data and environmental condition data through drones or ground sensors and extract key features from the collected spectral data and calculate relevant features based on crop growth stage and environmental conditions;

[0018] Based on historical data, a relationship model between spectral bands and crop growth stages and environmental conditions is established, and the relationship model is used to determine which spectral bands are most sensitive to specific crop states, and the current crop growth stage data and environmental condition data are input into the spectral selection model, and the spectral selection model is used to calculate the most suitable spectral bands under the current growth stage and environmental conditions;

[0019] The spectral sensor settings of the drone are adjusted according to the calculation results so that it can select and use the most suitable spectral band in data collection.

[0020] Optionally, the specific steps for calculating the most suitable spectral band under the current growth stage and environmental conditions are:

[0021] Calculate the spectral signal S i (λ) and predicted value M i The error between them is normalized and the error is weighted averaged, where the environmental adjustment factor To balance the environmental impact, the spectral band λ that minimizes the weighted error is selected as the optimal band, and its expression is: Among them, Optimal_Wavelength is the most suitable spectral band under the current crop growth stage and environmental conditions, λ is the wavelength of the spectral band, N is the number of spectral bands in the historical data, S i (λ) is the spectral signal intensity of the i-th spectral band at wavelength λ, M i The model predicts the spectral signal intensity for the i-th spectral band, is the variance of the spectral signal intensity of the i-th spectral band, β is the environmental adjustment factor, γ is the attenuation rate of the environmental adjustment factor, and Δ is the difference index between the current environmental conditions and the crop growth stage.

[0022] Optionally, the steps of performing spectral frequency conversion on the collected spectral data are:

[0023] Segmenting the spectral data into short time windows suitable for Fourier transform and performing Fourier transform on each data segment to obtain a frequency domain representation of the spectral data, and finally analyzing the data in the frequency domain;

[0024] According to the results of analyzing the data in the frequency domain, the key frequency components are identified, and the frequency components most sensitive to the target features are selected according to the analysis results. Frequency filters are applied to enhance the frequency components of interest and an inverse Fourier transform is applied to the frequency-adjusted spectrum data to restore it to the time domain or the original spectrum domain.

[0025] Optionally, the nonlinear spectral response model is a fully connected neural network model. When establishing the fully connected neural network model, the input layer, hidden layer, and output layer are first established, wherein the number of neurons in the input layer is consistent with the characteristic dimension of the spectral data, the number of hidden layers is 2, and each layer contains 64 neurons, and the number of neurons in the output layer is the same as the number of prediction targets. For example, if it is necessary to predict the moisture and nitrogen content at the same time, the output layer should contain 2 neurons. The steps for training the fully connected neural network model are:

[0026] Collect spectral data after spectral frequency conversion and amplitude modulation processing and analyze the spectral characteristics in the spectral data, identify the key frequency components related to water and nitrogen content, and pre-process and integrate the obtained data to form a data set;

[0027] The data set is divided into a training set and a test set, and the training set is used to train the fully connected neural network model. First, the input data is propagated through the input layer, hidden layer, and output layer of the neural network, and processed by the activation function ReLU to finally obtain the output prediction value. Then, the mean square error is used as the loss function to measure the gap between the predicted value and the actual value. Its expression is: Subsequently, the back-propagation algorithm is used to calculate the gradient of the loss function for each parameter, and SGD is used to update the weights and bias parameters of the network to minimize the loss function. After updating the weights and bias parameters of the network, the K-fold cross-validation technique is used to evaluate the generalization ability of the model, and the test set is used to evaluate the final performance of the model. The obtained model performance is used to determine whether the model needs to be optimized. Finally, the trained and verified fully connected neural network is integrated into the agricultural monitoring system, and the moisture and nitrogen content of wheat is automatically predicted by inputting real-time collected spectral data.

[0028] Optionally, when optimizing the parameters of the fully connected neural network model in real time, first, obtain the real-time performance of the model and dynamically adjust the learning rate based on the obtained real-time performance; when the model convergence speed slows down or the performance stagnates, reduce the learning rate for more detailed optimization; finally, use Bayesian optimization technology to automatically adjust the batch size and regularization coefficient of the model during the model operation.

[0029] Optionally, the steps for analyzing the data processed by spectral frequency conversion and amplitude modulation and generating intelligent irrigation and fertilization strategies based on prediction are as follows:

[0030] The newly collected spectral data is input into the trained fully connected neural network model to generate the prediction results of the current moisture and nitrogen content and generate an intelligent irrigation strategy based on the predicted moisture content and an intelligent fertilization strategy based on the predicted nitrogen content;

[0031] The generated intelligent irrigation and fertilization strategies are applied to actual farmland management and implemented through automated system operations.

[0032] Optionally, when the amount of water in the intelligent irrigation strategy and the amount of fertilizer in the intelligent fertilization strategy are determined during the irrigation strategy, first, the newly collected spectral data is input into the trained fully connected neural network model to obtain the predicted moisture content W pred and predicted nitrogen content N pred , calculate and predict the moisture W pred With the current moisture W currentThe difference between To adjust irrigation, generate required water, calculate and predict nitrogen N pred With the current nitrogen N current The difference between the application environment adjustment factor To adjust the amount of fertilizer and generate the required amount of fertilizer, the expression for obtaining the amount of water for the intelligent irrigation strategy is: The expression for obtaining the amount of fertilizer for the intelligent fertilization strategy is: Among them, A w is the water adjustment factor, which is used to adjust the sensitivity of the irrigation strategy. n is the nitrogen adjustment coefficient, which is used to adjust the sensitivity of fertilization strategy, γ n is the nitrogen environmental adjustment factor, which controls the impact of environmental factors on nitrogen fertilization strategy, γ w is the water environment adjustment factor, which controls the impact of environmental factors on water irrigation strategy, δ n is the attenuation rate of the nitrogen environmental adjustment factor, controlling the impact intensity of environmental factors, δ w is the attenuation rate of the moisture environment adjustment factor, which controls the impact intensity of environmental factors, T env is the ambient temperature, which affects the adjustment as part of the environmental conditions.

[0033] After adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art. Of course, any product implementing the present invention does not necessarily need to achieve all the advantages described below at the same time:

[0034] The present invention uses a multi-spectral sensor carried by an unmanned aerial vehicle to monitor the moisture and nitrogen content of wheat in real time, and uses a dynamic spectrum selection module to automatically select the optimal spectral band for data collection to ensure that the most accurate crop status information is obtained. The spectral frequency conversion module and the adaptive spectral amplitude modulation module in the system adjust the frequency and amplitude of the spectral data in real time to further optimize the data quality and accuracy. Based on these high-quality data, the data analysis and decision support module generates accurate intelligent irrigation and fertilization strategies, automatically adjusts the water and nitrogen supply to meet the actual needs of wheat at different growth stages, thereby optimizing the growth conditions of crops, significantly improving the growth quality and yield of wheat, and promoting agricultural production to develop in a more intelligent and precise direction, reducing resource waste and improving economic benefits.

[0035] The specific implementation modes of the present invention are further described in detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described below are only some embodiments. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0037] Figure 1 This is the block diagram of the intelligent monitoring system for water and nitrogen in drip-irrigated wheat based on UAV multispectral inversion.

[0038] It should be noted that these drawings and textual descriptions are not intended to limit the conceptual scope of the present invention in any way, but are intended to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0039] The present invention will now be described in further detail with reference to the accompanying drawings.

[0040] See also Figure 1 As shown, in this embodiment, a drip irrigation wheat water and nitrogen intelligent monitoring system based on drone multispectral inversion is provided, including a drone platform, and a sensor with multispectral imaging capability is mounted on the drone platform; the drone is equipped with a multispectral sensor to obtain spectral data of wheat crops in different bands. Multispectral sensors generally include visible light, near infrared and short-wave infrared bands, and can provide rich crop growth information, such as chlorophyll content, water status and nitrogen nutrition level.

[0041] Dynamic spectrum selection module, which is used to automatically select the most suitable spectral band for data collection according to the growth stage of crops and environmental conditions;

[0042] A spectrum frequency conversion module, which is used to convert the collected spectrum data into spectrum frequency;

[0043] An adaptive spectral amplitude modulation module, which is used to dynamically adjust the amplitude of the spectral signal during the spectral data acquisition process and automatically adjust the spectral amplitude according to real-time data feedback;

[0044] A nonlinear spectral response model module, which is used to establish a fully connected neural network model between spectral signals and wheat moisture and nitrogen content and optimize the parameters of the fully connected neural network model in real time;

[0045] Data analysis and decision support module, which is used to analyze the data processed by spectral frequency conversion and amplitude modulation, and generate intelligent irrigation and fertilization strategies based on prediction;

[0046] The control unit is used to coordinate the operation of the spectrum frequency conversion device, the adaptive spectrum amplitude modulation system and the dynamic spectrum selection module.

[0047] It should be noted that the present invention uses a multi-spectral sensor carried by an unmanned aerial vehicle to monitor the moisture and nitrogen content of wheat in real time, and uses a dynamic spectrum selection module to automatically select the optimal spectral band for data collection to ensure the most accurate crop status information. The spectral frequency conversion module and the adaptive spectral amplitude modulation module in the system adjust the frequency and amplitude of the spectral data in real time to further optimize the data quality and accuracy. Based on these high-quality data, the data analysis and decision support module generates accurate intelligent irrigation and fertilization strategies, automatically adjusts the water and nitrogen supply to meet the actual needs of wheat at different growth stages, thereby optimizing the growth conditions of crops, significantly improving the growth quality and yield of wheat, and promoting agricultural production to develop in a more intelligent and precise direction, reducing resource waste and improving economic benefits.

[0048] In this embodiment, the data analysis and decision support module includes a data fusion unit and a user interface module. The data fusion unit is used to comprehensively analyze the spectral data and other related data and generate an optimized irrigation and fertilization plan based on the analysis results. The user interface module is used to display real-time monitoring data, analysis results and optimization suggestions, and support users in making decisions. Through comprehensive analysis, the data fusion unit can generate an irrigation and fertilization plan that better meets actual needs. This can ensure that crops obtain the most appropriate resource supply under different growth stages and environmental conditions, thereby improving crop growth quality and yield. The user interface module displays real-time monitoring data and analysis results, so that users can clearly understand the current status of crops and the effects of management measures. This transparent operation can enhance users' trust and reliance on the system.

[0049] In this embodiment, the steps of automatically selecting the most suitable spectral band for data collection according to the growth stage and environmental conditions of the crop are:

[0050] The system collects crop growth stage data and environmental condition data through drones or ground sensors, extracts key features from the collected spectral data, and calculates relevant features based on the crop growth stage and environmental conditions; by extracting key features from spectral data, the system can focus on features that have a significant impact on the crop growth stage and environmental conditions, thereby improving the relevance and accuracy of the data. Based on the relationship model between spectral bands and crop status, the system can identify the spectral bands that are most sensitive to specific crop status, ensuring that the collected data reflects the crop status most accurately.

[0051] Based on historical data, a relationship model between spectral bands and crop growth stages and environmental conditions is established, and the relationship model is used to determine which spectral bands are most sensitive to specific crop states, and the current crop growth stage data and environmental condition data are input into the spectral selection model, and the spectral selection model is used to calculate the most suitable spectral bands under the current growth stage and environmental conditions; based on real-time crop growth stage data and environmental condition data, the most suitable spectral bands are calculated. This dynamic adaptability ensures that the data collection process always matches the actual conditions, thereby improving the monitoring effect.

[0052] By adjusting the spectral sensor settings of the drone based on the calculation results, the system is able to automatically select and use the most suitable spectral band in data collection. This reduces manual intervention and trial and error processes, and improves the efficiency and accuracy of data collection.

[0053] In this embodiment, the specific steps of calculating the most suitable spectral band under the current growth stage and environmental conditions are:

[0054] Calculate the spectral signal S i (λ) and predicted value M i The error between them is normalized and the error is weighted averaged, where the environmental adjustment factor To balance the environmental impact, the spectral band λ that minimizes the weighted error is selected as the optimal band, and its expression is: Among them, Optimal_Wavelength is the most suitable spectral band under the current crop growth stage and environmental conditions, λ is the wavelength of the spectral band, N is the number of spectral bands in the historical data, S i (λ) is the spectral signal intensity of the i-th spectral band at wavelength λ, M i The model predicts the spectral signal intensity for the i-th spectral band, is the variance of the spectral signal intensity of the i-th spectral band, β is the environmental adjustment factor, γ is the attenuation rate of the environmental adjustment factor, and Δ is the difference index between the current environmental conditions and the crop growth stage. By calculating the error between the spectral signal and the predicted value and standardizing it, the weighted average can select the spectral band that minimizes the weighted error. This ensures that under the current crop growth stage and environmental conditions, the selected spectral band reflects the crop status most accurately. The introduction of environmental adjustment factors to balance environmental influences can adjust the spectral signal according to changes in actual environmental conditions. Standardizing the error can eliminate the scale effect of the data, make the errors of different bands comparable, and improve the reliability and accuracy of error analysis.

[0055] In this embodiment, the steps of converting the collected spectral data into spectral frequency are as follows:

[0056] Segmenting the spectral data into short time windows suitable for Fourier transform and performing Fourier transform on each data segment to obtain a frequency domain representation of the spectral data, and finally analyzing the data in the frequency domain;

[0057] According to the results of analyzing the data in the frequency domain, the key frequency components are identified, and the frequency components most sensitive to the target features are selected according to the analysis results. Frequency filters are applied to enhance the frequency components of interest and an inverse Fourier transform is applied to the frequency-adjusted spectrum data to restore it to the time domain or the original spectrum domain.

[0058] In this embodiment, the nonlinear spectral response model is a fully connected neural network model. When the fully connected neural network model is established, the input layer, the hidden layer, and the output layer are first established, wherein the number of neurons in the input layer is consistent with the characteristic dimension of the spectral data, the number of hidden layers is 2 layers, and each layer contains 64 neurons, and the number of neurons in the output layer is the same as the number of prediction targets. For example, if it is necessary to predict the water content and nitrogen content at the same time, the output layer should contain 2 neurons. The steps of training the fully connected neural network model are:

[0059] Collect spectral data after spectral frequency conversion and amplitude modulation processing and analyze the spectral characteristics in the spectral data, identify the key frequency components related to water and nitrogen content, and pre-process and integrate the obtained data to form a data set;

[0060] The data set is divided into a training set and a test set, and the training set is used to train the fully connected neural network model. First, the input data is propagated through the input layer, hidden layer, and output layer of the neural network, and processed by the activation function ReLU to finally obtain the output prediction value. Then, the mean square error is used as the loss function to measure the gap between the predicted value and the actual value. Its expression is: Subsequently, the back-propagation algorithm is used to calculate the gradient of the loss function for each parameter, and SGD is used to update the weights and bias parameters of the network to minimize the loss function. After updating the weights and bias parameters of the network, the K-fold cross-validation technique is used to evaluate the generalization ability of the model, and the test set is used to evaluate the final performance of the model. The obtained model performance is used to determine whether the model needs to be optimized. Finally, the trained and verified fully connected neural network is integrated into the agricultural monitoring system, and the moisture and nitrogen content of wheat is automatically predicted by inputting real-time collected spectral data.

[0061] In this embodiment, when optimizing the parameters of the fully connected neural network model in real time, first, the real-time performance of the model is obtained and the learning rate is dynamically adjusted according to the obtained real-time performance. When the model convergence speed slows down or the performance stagnates, the learning rate is reduced for more detailed optimization. Finally, the Bayesian optimization technology is used to automatically adjust the batch size and regularization coefficient of the model during the model operation.

[0062] In this embodiment, the steps of analyzing the data processed by spectral frequency conversion and amplitude modulation and generating a prediction-based intelligent irrigation and fertilization strategy are as follows:

[0063] The newly collected spectral data is input into the trained fully connected neural network model to generate the prediction results of the current moisture and nitrogen content and generate an intelligent irrigation strategy based on the predicted moisture content and an intelligent fertilization strategy based on the predicted nitrogen content;

[0064] The generated intelligent irrigation and fertilization strategies are applied to actual farmland management and implemented through automated system operations.

[0065] By analyzing the data after spectral frequency conversion and amplitude modulation processing, the key frequency components related to moisture and nitrogen content can be identified, ensuring that the fully connected neural network model can capture the most critical features for crop status and improve the accuracy of prediction. By selecting a fully connected neural network model and training it with historical data, the complex relationship between spectral features and moisture and nitrogen content can be efficiently learned. The fully connected neural network model has strong feature extraction and pattern recognition capabilities, thus providing more accurate prediction results.

[0066] The newly collected spectral data is input into the trained fully connected neural network model to generate the prediction results of the current water and nitrogen content and generate intelligent irrigation strategies based on the predicted water content and intelligent fertilization strategies based on the predicted nitrogen content. Based on the prediction results, intelligent irrigation and fertilization strategies can accurately adjust the water and nitrogen supply and optimize crop growth conditions. This precise adjustment can effectively improve the health and yield of crops while reducing resource waste.

[0067] Apply the generated intelligent irrigation and fertilization strategies to actual farmland management and implement them through automated system operations. Applying the generated strategies to actual farmland management and implementing them through automated systems can improve management efficiency and consistency, reduce manual intervention, and ensure accurate implementation of strategies.

[0068] In this embodiment, when the amount of water in the intelligent irrigation strategy and the amount of fertilizer in the intelligent fertilization strategy are determined during the irrigation strategy, first, the newly collected spectral data is input into the trained fully connected neural network model to obtain the predicted moisture content W. pred and predicted nitrogen content Npred , calculate and predict the moisture W pred With the current moisture W current The difference between To adjust irrigation, generate required water, calculate and predict nitrogen N pred With the current nitrogen N current The difference between the application environment adjustment factor To adjust the amount of fertilizer and generate the required amount of fertilizer, the expression for obtaining the amount of water for the intelligent irrigation strategy is: The expression for obtaining the amount of fertilizer for the intelligent fertilization strategy is: Among them, A w is the water adjustment factor, which is used to adjust the sensitivity of the irrigation strategy. n is the nitrogen adjustment coefficient, which is used to adjust the sensitivity of fertilization strategy, γ n is the nitrogen environmental adjustment factor, which controls the impact of environmental factors on nitrogen fertilization strategy, γ w is the water environment adjustment factor, which controls the impact of environmental factors on water irrigation strategy, δ n is the attenuation rate of the nitrogen environmental adjustment factor, controlling the impact intensity of environmental factors, δ w is the attenuation rate of the moisture environment adjustment factor, which controls the impact intensity of environmental factors, T env The ambient temperature is used as part of the environmental conditions to influence the adjustment. By calculating the difference between the predicted water and nitrogen content and the current actual value, the actual irrigation and fertilization amount can be accurately adjusted so that the supply of water and fertilizer is more in line with the actual needs of the crop. Calculating and generating the required amount of water and fertilizer helps to optimize the use of resources, avoid over-application and waste, and ensure efficient use of resources. By controlling the influence of environmental factors, the system can adapt to different environmental conditions, effectively adjust the irrigation and fertilization strategies, and improve the adaptability and accuracy of management.

[0069] The present invention is not limited to the above-mentioned embodiments. Anyone should be aware that any structural changes made under the enlightenment of the present invention, and any technical solutions that are the same or similar to the present invention, fall within the protection scope of the present invention. The technology, shape, and structural parts not described in detail in the present invention are all well-known technologies.

Claims

1. An intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles, characterized in that: include: UAV platform, where the UAV platform is equipped with a sensor with multispectral imaging capability; Dynamic spectrum selection module, which is used to automatically select suitable spectral bands for data collection according to the growth stage of crops and environmental conditions; A spectrum frequency conversion module, which is used to convert the collected spectrum data into spectrum frequency; An adaptive spectral amplitude modulation module, which is used to dynamically adjust the amplitude of the spectral signal during the spectral data acquisition process and automatically adjust the spectral amplitude according to real-time data feedback; A nonlinear spectral response model module, which is used to establish a nonlinear spectral response model between spectral signals and wheat moisture and nitrogen content and optimize the parameters of the nonlinear spectral response model in real time; Data analysis and decision support module, which is used to analyze the data processed by spectral frequency conversion and amplitude modulation, and generate intelligent irrigation and fertilization strategies based on prediction; The control unit is used to coordinate the operation of the spectrum frequency conversion device, the adaptive spectrum amplitude modulation system and the dynamic spectrum selection module.

2. According to claim 1, a drip irrigation wheat water and nitrogen intelligent monitoring system based on drone multispectral inversion is characterized in that: The data analysis and decision support module includes a data fusion unit and a user interface module. The data fusion unit is used to comprehensively analyze the spectral data and other related data and generate optimized irrigation and fertilization plans based on the analysis results. The user interface module is used to display real-time monitoring data, analysis results and optimization suggestions, and support users in making decisions.

3. According to claim 1, the intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles is characterized in that: The steps to automatically select spectral bands for data collection based on crop growth stage and environmental conditions are: Collect crop growth stage data and environmental condition data through drones or ground sensors and extract key features from the collected spectral data and calculate relevant features based on crop growth stage and environmental conditions; Based on historical data, a relationship model between spectral bands and crop growth stages and environmental conditions is established, and the relationship model is used to determine which spectral bands are sensitive to specific crop states, and the current crop growth stage data and environmental condition data are input into the spectral selection model, and the spectral selection model is used to calculate the appropriate spectral bands under the current growth stage and environmental conditions; The spectral sensor settings of the UAV are adjusted according to the calculation results so that it can select and use appropriate spectral bands in data collection.

4. According to claim 1, the intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles is characterized in that: The specific steps to calculate the appropriate spectral band under the current growth stage and environmental conditions are: Calculate the spectral signal S i (λ) and predicted value M i The error between them is normalized and the error is weighted averaged, where the environmental adjustment factor To balance the environmental impact, the spectral band λ that makes the weighted error small is selected as the band, and its expression is: Among them, Optimal_Wavelength is the spectral band under the current crop growth stage and environmental conditions, λ is the wavelength of the spectral band, N is the number of spectral bands in the historical data, S i (λ) is the spectral signal intensity of the i-th spectral band at wavelength λ, M i The model predicts the spectral signal intensity for the i-th spectral band, is the variance of the spectral signal intensity of the i-th spectral band, β is the environmental adjustment factor, γ is the attenuation rate of the environmental adjustment factor, and Δ is the difference index between the current environmental conditions and the crop growth stage.

5. The intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles according to claim 1 is characterized in that: The steps for converting the collected spectral data into spectral frequency are as follows: Segmenting the spectral data into short time windows suitable for Fourier transform and performing Fourier transform on each data segment to obtain a frequency domain representation of the spectral data, and finally analyzing the data in the frequency domain; According to the results of analyzing the data in the frequency domain, the key frequency components are identified, and the frequency components sensitive to the target features are selected according to the analysis results. Frequency filters are applied to enhance the frequency components of interest and an inverse Fourier transform is applied to the frequency-adjusted spectrum data to restore it to the time domain or the original spectrum domain.

6. The intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles according to claim 1 is characterized in that: The nonlinear spectral response model is a fully connected neural network model. When establishing the fully connected neural network model, the input layer, hidden layer, and output layer are first established. The number of neurons in the input layer is consistent with the characteristic dimension of the spectral data, the number of hidden layers is 2, and each layer contains 64 neurons. The number of neurons in the output layer is the same as the number of predicted targets. The steps for training the fully connected neural network model are: Collect spectral data after spectral frequency conversion and amplitude modulation processing and analyze the spectral characteristics in the spectral data, identify the key frequency components related to water and nitrogen content, and pre-process and integrate the obtained data to form a data set; The data set is divided into a training set and a test set, and the training set is used to train the fully connected neural network model. First, the input data is propagated through the input layer, hidden layer, and output layer of the neural network, and processed by the activation function ReLU to finally obtain the output prediction value. Then, the mean square error is used as the loss function to measure the gap between the predicted value and the actual value. Its expression is: Subsequently, the back-propagation algorithm is used to calculate the gradient of the loss function for each parameter, and SGD is used to update the weights and bias parameters of the network to minimize the loss function. After updating the weights and bias parameters of the network, the K-fold cross-validation technique is used to evaluate the generalization ability of the model, and the test set is used to evaluate the final performance of the model. The obtained model performance is used to determine whether the model needs to be optimized. Finally, the trained and verified fully connected neural network is integrated into the agricultural monitoring system, and the moisture and nitrogen content of wheat is automatically predicted by inputting real-time collected spectral data.

7. The intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles according to claim 6 is characterized in that: When optimizing the parameters of the nonlinear spectral response model in real time, first, the real-time performance of the model is obtained and the learning rate is dynamically adjusted according to the obtained real-time performance. When the model convergence speed slows down or the performance stagnates, the learning rate is reduced for optimization. Finally, the Bayesian optimization technology is used to automatically adjust the batch size and regularization coefficient of the model during the model operation.

8. The intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles according to claim 7 is characterized in that: The steps to analyze the data processed by spectral frequency conversion and amplitude modulation and generate intelligent irrigation and fertilization strategies based on prediction are as follows: The newly collected spectral data is input into the trained fully connected neural network model to generate the prediction results of the current moisture and nitrogen content and generate an intelligent irrigation strategy based on the predicted moisture content and an intelligent fertilization strategy based on the predicted nitrogen content; The generated intelligent irrigation and fertilization strategies are applied to actual farmland management and implemented through automated system operations.

9. The intelligent monitoring system for water and nitrogen in drip irrigation wheat based on multi-spectral inversion of unmanned aerial vehicles according to claim 7 is characterized in that: When the amount of water in the intelligent irrigation strategy and the amount of fertilizer in the intelligent fertilization strategy are determined during the irrigation strategy, first, the newly collected spectral data is input into the trained fully connected neural network model to obtain the predicted moisture content W. pred and predicted nitrogen content N pred , calculate and predict the moisture W pred With the current moisture W current The difference between To adjust irrigation, generate required water, calculate and predict nitrogen N pred With the current nitrogen N current The difference between the application environment adjustment factor To adjust the amount of fertilizer and generate the required amount of fertilizer, the expression for obtaining the amount of water for the intelligent irrigation strategy is: The expression for obtaining the amount of fertilizer for the intelligent fertilization strategy is: Among them, A w is the water adjustment factor, which is used to adjust the sensitivity of the irrigation strategy. n is the nitrogen adjustment coefficient, which is used to adjust the sensitivity of fertilization strategy, γ n is the nitrogen environmental adjustment factor, which controls the impact of environmental factors on nitrogen fertilization strategy, γ w is the water environment adjustment factor, which controls the impact of environmental factors on water irrigation strategy, δ n is the attenuation rate of the nitrogen environmental adjustment factor, controlling the impact intensity of environmental factors, δ w is the attenuation rate of the moisture environment adjustment factor, which controls the impact intensity of environmental factors, T env is the ambient temperature, which affects the adjustment as part of the environmental conditions.

Images