A remote monitoring and management rice seedling irrigation and plant protection system, computer equipment and medium

Abstract

The present invention discloses a remote monitoring and management system for rice seedling irrigation and plant protection, computer equipment and medium, relating to the technical field of rice seedling irrigation, comprising a sensing acquisition module, an edge processing module, a data fusion module, a prediction modeling module, an execution control module, a user interaction module and an energy security component; the sensing acquisition module is used to collect soil moisture, temperature, pH value, air temperature and humidity, light intensity, wind speed and direction, rainfall and crop image information through multiple types of sensors deployed in the rice seedling field, and output raw environmental data; the edge processing module is connected to the sensing acquisition module, and is used to preprocess the raw environmental data, generate preprocessed data, and upload it to a cloud server via the NB-IoT protocol; the data fusion module is deployed in the cloud server, and is used to receive the preprocessed data from the edge processing module.

Description

Technical Field

[0001] The present invention relates to the technical field of rice seedling raising and irrigation, in particular to a remote monitoring and management rice seedling raising and irrigation plant protection system, computer equipment and medium. Background Art

[0002] Rice seedling irrigation refers to the process of managing and regulating the water content of the rice growing environment during the initial stages of rice planting, from seed germination to seedling growth. This period is crucial for ensuring healthy rice growth, as proper water supply promotes root development, enhances seedling resistance, and lays the foundation for subsequent tillering and heading. Therefore, utilizing advanced technologies to improve the intelligence and safety of rice seedling irrigation has become a pressing issue.

[0003] In the field of rice seedling irrigation, existing agricultural forecasting models cannot capture complex temporal relationships well, especially when facing uncertain factors such as climate change. The reliability of the forecast results is low, and traditional irrigation methods mostly rely on manual experience, which can easily lead to excessive or insufficient use of water resources and pesticides, which not only increases costs but also easily causes environmental pollution. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a remote monitoring and management rice seedling irrigation and plant protection system, computer equipment and medium to solve the problem that existing agricultural prediction models cannot well capture complex time series relationships, especially when facing uncertain factors such as climate change. The reliability of the prediction results is low, and traditional irrigation methods mostly rely on manual experience, which easily leads to excessive or insufficient use of water resources and pesticides, which not only increases costs but also easily causes environmental pollution.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a remote monitoring and management system for rice seedling irrigation and plant protection, comprising:

[0008] Sensing acquisition module, edge processing module, data fusion module, prediction modeling module, execution control module, user interaction module and energy security components;

[0009] The sensor collection module is used to collect soil moisture, temperature, pH value, air temperature and humidity, light intensity, wind speed and direction, rainfall, and crop image information through multiple types of sensors deployed in the rice seedling field, and output raw environmental data;

[0010] The edge processing module is connected to the sensor collection module and is used to pre-process the raw environmental data, generate pre-processed data, and upload it to the cloud server via the NB-IoT protocol;

[0011] The data fusion module is deployed in the cloud server and is used to receive the pre-processed data from the edge processing module, integrate and analyze the multi-source heterogeneous data using a dynamic weighted fusion algorithm, and generate a comprehensive farmland environmental state vector;

[0012] The prediction modeling module is connected to the data fusion module and is used to build an adaptive prediction model based on the farmland comprehensive environmental state vector and future weather forecast data, and output a prediction result;

[0013] The execution control module is connected to the prediction modeling module and is used to drive automatic drip irrigation and sprinkler irrigation equipment, GPS-guided drone spraying devices and water pump control equipment to perform local irrigation and plant protection operations based on the prediction results;

[0014] The user interaction module is in bidirectional communication with the prediction modeling module and the execution control module, and is used to provide a web and mobile terminal access interface, display real-time monitoring data, push decision suggestions, send alarm notifications, record execution logs, and support remote manual user intervention in the execution control module;

[0015] The energy security component is electrically connected to the sensing acquisition module, edge processing module and execution control module respectively, and is used to provide solar energy + battery hybrid power supply.

[0016] As a preferred solution of the remote monitoring and management rice seedling irrigation and plant protection system of the present invention, the raw environmental data is preprocessed to generate preprocessed data, specifically in the following steps:

[0017] A low-pass filter is used to filter and denoise the continuous numerical raw data collected by the soil moisture sensor to remove high-frequency noise interference and obtain filtered soil moisture data.

[0018] The sliding average filtering algorithm is used to smooth the continuous numerical raw data collected by the soil temperature sensor to eliminate random fluctuations and obtain filtered soil temperature data.

[0019] The limiting filtering method is used to remove abnormal values ​​from the continuous numerical raw data collected by the soil pH sensor, limiting the data range to a reasonable range, and obtaining the filtered soil pH data;

[0020] The normalization formula is used to convert the two-dimensional numerical raw data collected by the air temperature and humidity sensor into a standardized format to obtain the standardized air temperature and humidity data;

[0021] The linear transformation method is used to standardize the continuous numerical raw data collected by the light intensity sensor to obtain the standardized light intensity data;

[0022] The coordinate transformation formula is used to calculate the horizontal wind speed component of the continuous numerical raw data collected by the three-dimensional wind speed and direction sensor to obtain the standardized wind speed data. The expression is:

[0023] ;

[0024] in, is the normalized wind speed data, , are the original data of the horizontal component of wind speed, is the minimum value of the east-west wind speed component in historical records. is the maximum value of the east-west wind speed component in historical records. is the minimum value of the north-south wind speed component in historical records. It is the maximum value of the north-south wind speed component in historical records;

[0025] The linear scaling method is used to standardize the continuous numerical raw data collected by the tipping bucket rain gauge, and the precipitation data is mapped to the interval [0,1] to obtain the standardized rainfall data.

[0026] For the discrete numerical raw data collected by agricultural cameras, the pixel intensity ratio calculation formula is directly used for normalization processing to obtain the standardized crop image data;

[0027] After completing the filtering, denoising and standardization of the above sensor data, a timestamp is added to all the standardized data to generate a preprocessed data set containing time information.

[0028] As a preferred solution of the remote monitoring and management rice seedling irrigation and plant protection system of the present invention, the steps of receiving pre-processed data from the edge processing module, integrating and analyzing multi-source heterogeneous data using a dynamic weighted fusion algorithm, and generating a comprehensive farmland environment state vector are as follows:

[0029] A cloud server is used to receive the pre-processed data set uploaded by the edge processing module;

[0030] The dynamic weight calculation formula is used to assign weights to the sensor data to reflect the importance differences of different sensor data under the current environmental conditions, and the weighted coefficient is obtained. The expression is:

[0031] ;

[0032] in, For the The weight coefficient of sensor data, For the The actual measured value of the sensor data, For the The ideal optimal value of sensor data, is an adjustment parameter used to control the concentration of weight distribution. and Represent the actual measured values ​​and ideal optimal values ​​of the data of other sensors respectively;

[0033] The normalized weighted fusion formula is used to integrate and analyze the weighted sensor data to generate the farmland comprehensive environmental state vector, which is expressed as:

[0034] ;

[0035] in, is the comprehensive environmental state vector of farmland, For the The weight coefficient of sensor data, For the The normalized value of sensor data;

[0036] The generated farmland comprehensive environmental state vector Serves as the data input basis for subsequent predictive modeling modules, driving the system to perform irrigation and plant protection decision analysis;

[0037] For crop image data, image feature extraction algorithm is used to process it, identify the health status of crops, and append the extracted image feature vector to the farmland comprehensive environmental state vector to form an enhanced comprehensive environmental state vector. , the expression is:

[0038] ;

[0039] in, is the enhanced comprehensive environment state vector, is the image correction factor, is the feature vector extracted from the crop image.

[0040] As a preferred solution of the remote monitoring and management rice seedling irrigation and plant protection system of the present invention, wherein: based on the farmland comprehensive environmental state vector and future weather forecast data, an adaptive prediction model is constructed to output the prediction result, and the specific steps are as follows:

[0041] The cloud server receives the enhanced comprehensive environmental state vector generated by the data fusion module and the weather forecast data for the next 24 hours obtained from the external weather service API. ;

[0042] The sliding window method is used to analyze the historical environment state vector and corresponding weather forecast data Divide the time series to form a training sample set;

[0043] The long short-term memory network LSTM is used as the basic model to train the training sample set and learn the environment state vector and weather forecast data The dynamic relationship between them is used to build an adaptive prediction model, which is expressed as:

[0044] ;

[0045] in, For the The hidden state of time steps, For the The enhanced comprehensive environment state vector of time steps, For the Weather forecast data for time steps, is the parameter set of the LSTM model;

[0046] Use the fully connected layer to hide the last layer of the LSTM network Perform linear transformation to obtain the predicted output result, the expression is:

[0047] ;

[0048] in, is the predicted output vector, is the weight matrix of the fully connected layer, is the bias vector of the fully connected layer, is the hidden state of the last time step;

[0049] Adopting an adaptive learning rate adjustment strategy, the learning rate is automatically adjusted during each iteration to accelerate convergence and improve model accuracy. The expression is:

[0050] ;

[0051] in, are the updated model parameters, is the initial learning rate, and are the first-order moment estimate and the second-order moment estimate, respectively. To prevent division by zero for small constants, For the The model parameter values ​​for the iteration;

[0052] Apply the trained adaptive prediction model to real-time data and generate prediction results based on the current enhanced integrated environmental state vector and future weather forecast data;

[0053] The prediction results include optimal irrigation timing, water requirements, fertilization timing and dosage recommendations, and plant protection intervention priorities.

[0054] As a preferred solution of the remote monitoring and management rice seedling irrigation and plant protection system of the present invention, wherein: the automatic drip irrigation and sprinkler irrigation equipment, GPS-guided drone spraying device and water pump control equipment are driven according to the prediction results to perform local irrigation and plant protection operations, and the specific steps are as follows:

[0055] Use the decision engine to analyze and classify the prediction results and generate a specific execution instruction set;

[0056] The NB-IoT communication protocol is used to transmit the execution instruction set to each execution control device in the field;

[0057] For automatic drip irrigation equipment: A soil moisture sensor is used to monitor soil moisture content in real time. When the current soil moisture is detected to be lower than the set threshold, the drip irrigation equipment is triggered to start, and precise irrigation is carried out according to the predicted water demand.

[0058] When the set stop condition is reached, the drip irrigation equipment is turned off. The expression is:

[0059] ;

[0060] in, is the current soil moisture, is the minimum soil moisture threshold set, is the predicted water demand;

[0061] For sprinkler irrigation equipment: The rainfall forecast value in the weather forecast data is used in combination with the current environmental state vector to determine whether the sprinkler irrigation equipment needs to be activated. If there is no effective rainfall in the next 24 hours and the soil moisture is below the set threshold, the sprinkler irrigation equipment is activated and irrigation is carried out according to the water demand;

[0062] When the set stop condition is reached, the sprinkler equipment is turned off. The expression is:

[0063] ;

[0064] in, The rainfall forecast for the next 24 hours;

[0065] For GPS-guided drone spraying devices: A high-precision GPS positioning system is used to obtain farmland map zoning information. Combined with the plant protection intervention priority (PCP) in the predicted results, the specific areas to be sprayed are determined. The required pesticide dosage is calculated based on the crop growth stage and pest and disease risk level. The drone sprayer then precisely applies the pesticide. After completing the spraying task in the designated area, the drone returns to the starting point and records the operation log.

[0066] For water pump control equipment: a water level sensor is used to monitor the water level in the reservoir in real time. When the water level is detected to be lower than the set threshold, the water pump is started to replenish the water source. When the water level returns to the set safe water level, the water pump is shut down. Combined with the water demand in the predicted results, the operating frequency of the water pump is dynamically adjusted to meet irrigation needs.

[0067] As a preferred solution for the remote monitoring and management of the rice seedling irrigation and plant protection system of the present invention, wherein: the web and mobile terminal access interfaces are provided to display real-time monitoring data, push decision suggestions, send alarm notifications, record execution logs, and support users to remotely and manually intervene in the execution control module. The specific steps are as follows:

[0068] Use the front-end development framework React to build the web and mobile application interface UI, and achieve two-way communication with the cloud server through the API interface;

[0069] The WebSocket protocol is used to achieve real-time data transmission, and the real-time monitoring data from the sensor acquisition module and the execution control module are transmitted to the Web and mobile applications;

[0070] Use the visual chart library ECharts to graphically display real-time monitoring data and generate intuitive environmental status trend charts and equipment operation status charts;

[0071] The message push mechanism is used to push the decision suggestions generated by the prediction modeling module to users. The expression is:

[0072] ;

[0073] in, A set of decision suggestions, The best irrigation time point is is the water requirement, For fertilizer timing and dosage recommendations, Prioritize plant protection interventions;

[0074] The alarm rule engine is used to analyze real-time monitoring data. When an abnormal situation is detected, an alarm notification is triggered and sent to the user via SMS or email.

[0075] As a preferred solution of the remote monitoring and management rice seedling irrigation and plant protection system of the present invention, the adaptive learning rate adjustment strategy is adopted to automatically adjust the learning rate during each iteration to accelerate convergence and improve model accuracy. The specific process is as follows:

[0076] Get the gradient of the loss function at the current time step, and calculate the loss function based on the training samples of the current batch through the back propagation algorithm Model parameters The gradient of is expressed as:

[0077] ;

[0078] in, is a The dimension of the model parameter value at the iteration The same vector, indicating that The partial derivative of the loss function with respect to the parameters at time steps is, represents the model parameters;

[0079] Calculate the first-order moment estimate of the current gradient, the expression is:

[0080] ;

[0081] in, is the first-order moment estimate of the previous time step, Control the retention ratio of historical gradients, is the contribution of the current gradient, Reflects the changing trend of the gradient;

[0082] Calculate the second-order moment estimate of the current gradient, the expression is:

[0083] ;

[0084] in, is the second-order moment estimate of the previous time step, Indicates that each element of the gradient vector is squared. Used to measure the degree of fluctuation of gradient changes, is the second-order moment attenuation coefficient;

[0085] Since the initial moment and Are all 0, there will be deviations in the first few rounds, and they need to be corrected unbiasedly. The expression is:

[0086] ;

[0087] in, is the first-order moment estimate after bias correction, is the bias-corrected second-order moment estimate, and are exponential power terms respectively.

[0088] As a preferred solution of the remote monitoring and management rice seedling irrigation and plant protection system of the present invention, wherein: the long short-term memory network LSTM is used as the basic model, and the loss function needs to be calculated during the training of the training sample set, including:

[0089] During the training process, the back-propagation through time algorithm is used to update the gradient of the LSTM model;

[0090] The input sequence is divided into time segments of fixed length, which are input into LSTM in sequence. The prediction error of each time step is calculated, and the error is back-propagated to the initial time step through the chain rule. The expression is:

[0091] ;

[0092] in, is the total loss function, is the loss function, is the true target value, is the model's predicted value.

[0093] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the remote monitoring and management of the rice seedling irrigation and plant protection system as described in the first aspect of the present invention is implemented.

[0094] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the remote monitoring and management of the rice seedling irrigation and plant protection system as described in the first aspect of the present invention is implemented.

[0095] The beneficial effects of the present invention are as follows: dynamic weight allocation is used to automatically adjust the importance of each sensor data according to different environmental conditions; weighted fusion is used to realize intelligent integration of multi-source heterogeneous data, thereby improving the comprehensiveness and accuracy of environmental perception; an image feature addition step is used to realize non-contact evaluation of crop growth status, thereby enhancing the system's ability to identify practical problems; a sliding window is used to divide the sample step to realize effective modeling of time series data and capture long-term dependencies; an LSTM modeling step is used to realize modeling and prediction of changing trends in complex agricultural environments, thereby improving the generalization ability and prediction accuracy of the prediction model and being able to adapt to agricultural production needs under different climatic conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0097] Figure 1 This is a flow chart of the remote monitoring and management of the rice seedling irrigation and plant protection system in Example 1. DETAILED DESCRIPTION

[0098] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0099] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0100] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0101] Example 1, with reference to Figure 1 , which is the first embodiment of the present invention, provides a remote monitoring and management rice seedling irrigation and plant protection system, comprising:

[0102] Sensing acquisition module, edge processing module, data fusion module, prediction modeling module, execution control module, user interaction module and energy security components;

[0103] The sensing acquisition module is used to collect soil moisture, temperature, pH value, air temperature and humidity, light intensity, wind speed and direction, rainfall, and crop image information through multiple types of sensors deployed in rice seedling fields, and output raw environmental data.

[0104] An edge processing module, connected to the sensor collection module, is used to pre-process the raw environmental data, generate pre-processed data, and upload it to the cloud server via the NB-IoT protocol;

[0105] Furthermore, a low-pass filter is used to filter and denoise the continuous numerical raw data collected by the soil moisture sensor to remove high-frequency noise interference and obtain filtered soil moisture data;

[0106] The sliding average filtering algorithm is used to smooth the continuous numerical raw data collected by the soil temperature sensor to eliminate random fluctuations and obtain filtered soil temperature data.

[0107] The limiting filtering method is used to remove abnormal values ​​from the continuous numerical raw data collected by the soil pH sensor, limiting the data range to a reasonable range, and obtaining the filtered soil pH data;

[0108] The normalization formula is used to convert the two-dimensional numerical raw data collected by the air temperature and humidity sensor into a standardized format to obtain the standardized air temperature and humidity data;

[0109] The linear transformation method is used to standardize the continuous numerical raw data collected by the light intensity sensor to obtain the standardized light intensity data;

[0110] The coordinate transformation formula is used to calculate the horizontal wind speed component of the continuous numerical raw data collected by the three-dimensional wind speed and direction sensor to obtain the standardized wind speed data. The expression is:

[0111] ;

[0112] in, is the normalized wind speed data, , are the original data of the horizontal component of wind speed, is the minimum value of the east-west wind speed component in historical records. is the maximum value of the east-west wind speed component in historical records. is the minimum value of the north-south wind speed component in historical records. It is the maximum value of the north-south wind speed component in historical records;

[0113] The linear scaling method is used to standardize the continuous numerical raw data collected by the tipping bucket rain gauge, and the precipitation data is mapped to the interval [0,1] to obtain the standardized rainfall data.

[0114] For the discrete numerical raw data collected by agricultural cameras, the pixel intensity ratio calculation formula is directly used for normalization processing to obtain the standardized crop image data;

[0115] After completing the filtering, denoising and standardization of the above sensor data, add a timestamp to all the standardized data to generate a preprocessed data set containing time information;

[0116] It should be noted that filtering and denoising, standardized format conversion, and timestamp preprocessing operations are completed on the edge side, effectively reducing the communication burden of uploading raw data to the cloud, improving the system's response speed and energy efficiency, and using the NB-IoT protocol for remote transmission, ensuring stable, low-latency data upload to the cloud even in remote farmland areas, thereby enhancing the system's applicability and deployment flexibility.

[0117] The data fusion module is deployed in the cloud server and is used to receive pre-processed data from the edge processing module. It uses a dynamic weighted fusion algorithm to integrate and analyze multi-source heterogeneous data to generate a comprehensive farmland environmental state vector.

[0118] Furthermore, a cloud server is used to receive the pre-processed dataset uploaded by the edge processing module;

[0119] The dynamic weight calculation formula is used to assign weights to the sensor data to reflect the importance differences of different sensor data under the current environmental conditions, and the weighted coefficient is obtained. The expression is:

[0120] ;

[0121] in, For the The weight coefficient of sensor data, For the The actual measured value of the sensor data, For the The ideal optimal value of sensor data, is an adjustment parameter used to control the concentration of weight distribution. and Represent the actual measured values ​​and ideal optimal values ​​of the data of other sensors respectively;

[0122] The normalized weighted fusion formula is used to integrate and analyze the weighted sensor data to generate the farmland comprehensive environmental state vector, which is expressed as:

[0123] ;

[0124] in, is the comprehensive environmental state vector of farmland, For the The weight coefficient of sensor data, For the The normalized value of sensor data;

[0125] The generated farmland comprehensive environmental state vector Serves as the data input basis for subsequent predictive modeling modules, driving the system to perform irrigation and plant protection decision analysis;

[0126] For crop image data, image feature extraction algorithm is used to process it, identify the health status of crops, and append the extracted image feature vector to the farmland comprehensive environmental state vector to form an enhanced comprehensive environmental state vector. , the expression is:

[0127] ;

[0128] in, is the enhanced comprehensive environment state vector, is the image correction factor, is the feature vector extracted from the crop image;

[0129] It should be noted that the dynamic weighted fusion algorithm can adaptively adjust the weight of each sensor data according to the current environmental status, avoiding overall judgment bias caused by the abnormality or failure of a single sensor, thereby improving the accuracy and robustness of the comprehensive environmental state vector. In addition, the introduction of image feature vectors enables the system to not only reflect physical environmental parameters, but also evaluate crop health from a visual perspective, further improving the input quality and decision reliability of the prediction model.

[0130] The prediction modeling module is connected to the data fusion module and is used to build an adaptive prediction model based on the comprehensive farmland environmental state vector and future weather forecast data, and output the prediction results;

[0131] Furthermore, a cloud server is used to receive the enhanced comprehensive environmental state vector generated by the data fusion module and the weather forecast data for the next 24 hours obtained from the external weather service API. ;

[0132] The sliding window method is used to analyze the historical environment state vector and corresponding weather forecast data Divide the time series to form a training sample set;

[0133] The long short-term memory network LSTM is used as the basic model to train the training sample set and learn the environment state vector and weather forecast data The dynamic relationship between them is used to build an adaptive prediction model, which is expressed as:

[0134] ;

[0135] in, For the The hidden state of time steps, For the The enhanced comprehensive environment state vector of time steps, For the Weather forecast data for time steps, is the parameter set of the LSTM model;

[0136] Use the fully connected layer to hide the last layer of the LSTM network Perform linear transformation to obtain the predicted output result, the expression is:

[0137] ;

[0138] in, is the predicted output vector, is the weight matrix of the fully connected layer, is the bias vector of the fully connected layer, is the hidden state of the last time step;

[0139] Adopting an adaptive learning rate adjustment strategy, the learning rate is automatically adjusted during each iteration to accelerate convergence and improve model accuracy. The expression is:

[0140] ;

[0141] in, are the updated model parameters, is the initial learning rate, and are the first-order moment estimate and the second-order moment estimate, respectively. To prevent division by zero for small constants, For the The model parameter values ​​for the iteration;

[0142] Apply the trained adaptive prediction model to real-time data and generate prediction results based on the current enhanced integrated environmental state vector and future weather forecast data;

[0143] The prediction results include optimal irrigation timing, water requirements, fertilizer timing and dosage recommendations, and plant protection intervention priorities;

[0144] Adopting an adaptive learning rate adjustment strategy, the learning rate is automatically adjusted during each iteration to accelerate convergence and improve model accuracy. The specific process is as follows:

[0145] Get the gradient of the loss function at the current time step, and calculate the loss function based on the training samples of the current batch through the back propagation algorithm Model parameters The gradient of is expressed as:

[0146] ;

[0147] in, is a The dimension of the model parameter value at the iteration The same vector, indicating that The partial derivative of the loss function with respect to the parameters at time steps is, represents the model parameters;

[0148] Calculate the first-order moment estimate of the current gradient, the expression is:

[0149] ;

[0150] in, is the first-order moment estimate of the previous time step, Control the retention ratio of historical gradients, is the contribution of the current gradient, Reflects the changing trend of the gradient;

[0151] Calculate the second-order moment estimate of the current gradient, the expression is:

[0152] ;

[0153] in, is the second-order moment estimate of the previous time step, Indicates that each element of the gradient vector is squared. Used to measure the degree of fluctuation of gradient changes, is the second-order moment attenuation coefficient;

[0154] Since the initial moment and Are all 0, there will be deviations in the first few rounds, and they need to be corrected unbiasedly. The expression is:

[0155] ;

[0156] in, is the first-order moment estimate after bias correction, is the bias-corrected second-order moment estimate, and are the exponential power terms respectively;

[0157] Using the long short-term memory network LSTM as the basic model, the loss function needs to be calculated during the training of the training sample set, including:

[0158] During the training process, the back-propagation through time algorithm is used to update the gradient of the LSTM model;

[0159] The input sequence is divided into time segments of fixed length, which are input into LSTM in sequence. The prediction error of each time step is calculated, and the error is back-propagated to the initial time step through the chain rule. The expression is:

[0160] ;

[0161] in, is the total loss function, is the loss function, is the true target value, is the model prediction value, is the total number of time steps, i.e. the length of the sequence;

[0162] It should be noted that the LSTM-based adaptive forecasting model uses a sliding window approach to construct a training sample set, fully exploiting the temporal dependency between historical environmental conditions and future weather forecasts. Combined with the Adam optimizer's adaptive learning rate strategy, this significantly improves the model's convergence speed and generalization capabilities, making forecasts more aligned with practical farmland management needs. The forecasting model can be fine-tuned online based on different regions, seasons, and crop stages, enhancing its adaptability and practicality.

[0163] The execution control module is connected to the prediction modeling module and is used to drive automatic drip irrigation and sprinkler irrigation equipment, GPS-guided drone spraying devices, and water pump control equipment to perform local irrigation and plant protection operations based on the prediction results;

[0164] Furthermore, a decision engine is used to parse and classify the prediction results and generate a specific set of execution instructions;

[0165] The NB-IoT communication protocol is used to transmit the execution instruction set to each execution control device in the field;

[0166] For automatic drip irrigation equipment: A soil moisture sensor is used to monitor soil moisture content in real time. When the current soil moisture is detected to be lower than the set threshold, the drip irrigation equipment is triggered to start, and precise irrigation is carried out according to the predicted water demand.

[0167] When the set stop condition is reached, the drip irrigation equipment is turned off. The expression is:

[0168] ;

[0169] in, is the current soil moisture, is the minimum soil moisture threshold set, is the predicted water demand;

[0170] For sprinkler irrigation equipment: The rainfall forecast value in the weather forecast data is used in combination with the current environmental state vector to determine whether the sprinkler irrigation equipment needs to be activated. If there is no effective rainfall in the next 24 hours and the soil moisture is below the set threshold, the sprinkler irrigation equipment is activated and irrigation is carried out according to the water demand;

[0171] When the set stop condition is reached, the sprinkler equipment is turned off. The expression is:

[0172] ;

[0173] in, The rainfall forecast for the next 24 hours;

[0174] For GPS-guided drone spraying devices: A high-precision GPS positioning system is used to obtain farmland map zoning information. Combined with the plant protection intervention priority (PCP) in the predicted results, the specific areas to be sprayed are determined. The required pesticide dosage is calculated based on the crop growth stage and pest and disease risk level. The drone sprayer then precisely applies the pesticide. After completing the spraying task in the designated area, the drone returns to the starting point and records the operation log.

[0175] For water pump control equipment: A water level sensor is used to monitor the water level in the reservoir in real time. When the water level is detected to be below the set threshold, the water pump is started to replenish the water source. When the water level returns to the set safe water level, the water pump is shut down. Based on the water demand in the predicted results, the operating frequency of the water pump is dynamically adjusted to meet the irrigation needs.

[0176] It should be noted that automatic drip irrigation, sprinkler irrigation equipment and drone plant protection devices are precisely controlled based on prediction results and real-time environmental feedback, realizing closed-loop management from data-driven to behavioral output. By setting threshold trigger mechanisms and dynamic frequency adjustment strategies, it not only ensures the timeliness and effectiveness of irrigation and pesticide application, but also avoids resource waste and environmental pollution, which helps promote the deep integration of green agriculture and intelligent agricultural machinery.

[0177] The user interaction module has a two-way communication connection with the prediction modeling module and the execution control module. It is used to provide web and mobile access interfaces, display real-time monitoring data, push decision suggestions, send alarm notifications, record execution logs, and support users to remotely and manually intervene in the execution control module.

[0178] Furthermore, the front-end development framework React is used to build the web and mobile application interface UI, and two-way communication with the cloud server is achieved through the API interface;

[0179] The WebSocket protocol is used to achieve real-time data transmission, and the real-time monitoring data from the sensor acquisition module and the execution control module are transmitted to the Web and mobile applications;

[0180] Use the visual chart library ECharts to graphically display real-time monitoring data and generate intuitive environmental status trend charts and equipment operation status charts;

[0181] The message push mechanism is used to push the decision suggestions generated by the prediction modeling module to users. The expression is:

[0182] ;

[0183] in, A set of decision suggestions, The best irrigation time point is is the water requirement, For fertilizer timing and dosage recommendations, Prioritize plant protection interventions;

[0184] Use the alarm rule engine to analyze real-time monitoring data. When an abnormal situation is detected, an alarm notification is triggered and sent to the user via SMS or email.

[0185] It should be noted that the web and mobile access interfaces support remote access and operation on multiple platforms and terminals, meeting the needs of modern agricultural managers to grasp the status of farmland and perform interventions anytime and anywhere. Real-time data push, ECharts visualization and alarm notification mechanism are realized through WebSocket, which improves the transparency of the system and user participation. At the same time, it provides a traceable and auditable operation log system for the agricultural production process, enhancing the security and management efficiency of the system.

[0186] The energy security component is electrically connected to the sensor acquisition module, the edge processing module and the execution control module respectively, and is used to provide solar energy + battery hybrid power supply.

[0187] This embodiment also provides a computer device suitable for remote monitoring and management of rice seedling irrigation and plant protection systems, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement remote monitoring and management of rice seedling irrigation and plant protection systems as proposed in the above embodiments.

[0188] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0189] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, it implements the remote monitoring and management of the rice seedling irrigation and plant protection system as proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0190] In summary, the present invention realizes automatic adjustment of the importance of each sensor data according to different environmental conditions through dynamic weight allocation, realizes intelligent integration of multi-source heterogeneous data through weighted fusion, improves the comprehensiveness and accuracy of environmental perception, realizes non-contact evaluation of crop growth status through image feature addition step, enhances the system's ability to identify practical problems, realizes effective modeling of time series data and captures long-term dependencies through the sliding window sample division step, realizes modeling and prediction of complex agricultural environment change trends through LSTM modeling step, improves the generalization ability and prediction accuracy of the prediction model, and can adapt to agricultural production needs under different climatic conditions.

[0191] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A remote monitoring and management system for rice seedling irrigation and plant protection, characterized by: include: Sensing acquisition module, edge processing module, data fusion module, prediction modeling module, execution control module, user interaction module and energy security components; The sensor collection module is used to collect soil moisture, temperature, pH value, air temperature and humidity, light intensity, wind speed and direction, rainfall, and crop image information through multiple types of sensors deployed in the rice seedling field, and output raw environmental data; The edge processing module is connected to the sensor collection module and is used to pre-process the raw environmental data, generate pre-processed data, and upload it to the cloud server via the NB-IoT protocol; The data fusion module is deployed in the cloud server and is used to receive the pre-processed data from the edge processing module, integrate and analyze the multi-source heterogeneous data using a dynamic weighted fusion algorithm, and generate a comprehensive farmland environmental state vector; A cloud server is used to receive the pre-processed data set uploaded by the edge processing module; The dynamic weight calculation formula is used to assign weights to the sensor data to reflect the importance differences of different sensor data under the current environmental conditions, and the weighted coefficient is obtained. The expression is: ; in, For the The weight coefficient of sensor data, For the The actual measured value of the sensor data, For the The ideal optimal value of sensor data, is an adjustment parameter used to control the concentration of weight distribution. and Represent the actual measured values ​​and ideal optimal values ​​of the data of other sensors respectively; The normalized weighted fusion formula is used to integrate and analyze the weighted sensor data to generate the farmland comprehensive environmental state vector, which is expressed as: ; in, is the comprehensive environmental state vector of farmland, For the The weight coefficient of sensor data, For the The normalized value of sensor data; The generated farmland comprehensive environmental state vector Serves as the data input basis for subsequent predictive modeling modules, driving the system to perform irrigation and plant protection decision analysis; For crop image data, image feature extraction algorithm is used to process it, identify the health status of crops, and append the extracted image feature vector to the farmland comprehensive environmental state vector to form an enhanced comprehensive environmental state vector. , the expression is: ; in, is the enhanced comprehensive environment state vector, is the image correction factor, is the feature vector extracted from the crop image; The prediction modeling module is connected to the data fusion module and is used to build an adaptive prediction model based on the farmland comprehensive environmental state vector and future weather forecast data, and output a prediction result; The cloud server receives the enhanced comprehensive environmental state vector generated by the data fusion module and the weather forecast data for the next 24 hours obtained from the external weather service API. ; The sliding window method is used to analyze the historical environment state vector and corresponding weather forecast data Divide the time series to form a training sample set; The long short-term memory network LSTM is used as the basic model to train the training sample set and learn the environment state vector and weather forecast data The dynamic relationship between them is used to build an adaptive prediction model, which is expressed as: ； in, For the The hidden state of time steps, For the The enhanced comprehensive environment state vector of time steps, For the Weather forecast data for time steps, is the parameter set of the LSTM model; Use the fully connected layer to hide the last layer of the LSTM network Perform linear transformation to obtain the predicted output result, the expression is: ； in, is the predicted output vector, is the weight matrix of the fully connected layer, is the bias vector of the fully connected layer, is the hidden state of the last time step; Adopting an adaptive learning rate adjustment strategy, the learning rate is automatically adjusted during each iteration to accelerate convergence and improve model accuracy. The expression is: ; in, are the updated model parameters, is the initial learning rate, and are the first-order moment estimate and the second-order moment estimate, respectively. To prevent division by zero for small constants, For the The model parameter values ​​for the iteration; Apply the trained adaptive prediction model to real-time data and generate prediction results based on the current enhanced integrated environmental state vector and future weather forecast data; The prediction results include optimal irrigation timing, water requirements, fertilizer timing and dosage recommendations, and plant protection intervention priorities; The execution control module is connected to the prediction modeling module and is used to drive automatic drip irrigation and sprinkler irrigation equipment, GPS-guided drone spraying devices and water pump control equipment to perform local irrigation and plant protection operations based on the prediction results; Use the decision engine to analyze and classify the prediction results and generate a specific execution instruction set; The NB-IoT communication protocol is used to transmit the execution instruction set to each execution control device in the field; For automatic drip irrigation equipment: A soil moisture sensor is used to monitor soil moisture content in real time. When the current soil moisture is detected to be lower than the set threshold, the drip irrigation equipment is triggered to start, and precise irrigation is carried out according to the predicted water demand. When the set stop condition is reached, the drip irrigation equipment is turned off. The expression is: ； in, is the current soil moisture, is the minimum soil moisture threshold set, is the predicted water demand; For sprinkler irrigation equipment: The rainfall forecast value in the weather forecast data is used in combination with the current environmental state vector to determine whether the sprinkler irrigation equipment needs to be activated. If there is no effective rainfall in the next 24 hours and the soil moisture is below the set threshold, the sprinkler irrigation equipment is activated and irrigation is carried out according to the water demand; When the set stop condition is reached, the sprinkler equipment is turned off. The expression is: ； in, The rainfall forecast for the next 24 hours; For GPS-guided drone spraying devices: A high-precision GPS positioning system is used to obtain farmland map zoning information. Combined with the plant protection intervention priority (PCP) in the predicted results, the specific areas to be sprayed are determined. The required pesticide dosage is calculated based on the crop growth stage and pest and disease risk level. The drone sprayer then precisely applies the pesticide. After completing the spraying task in the designated area, the drone returns to the starting point and records the operation log. For water pump control equipment: A water level sensor is used to monitor the water level in the reservoir in real time. When the water level is detected to be below the set threshold, the water pump is started to replenish the water source. When the water level returns to the set safe water level, the water pump is shut down. Based on the water demand in the predicted results, the operating frequency of the water pump is dynamically adjusted to meet the irrigation needs. The user interaction module is in bidirectional communication with the prediction modeling module and the execution control module, and is used to provide a web and mobile terminal access interface, display real-time monitoring data, push decision suggestions, send alarm notifications, record execution logs, and support remote manual user intervention in the execution control module; The energy security component is electrically connected to the sensing acquisition module, edge processing module and execution control module respectively, and is used to provide solar energy + battery hybrid power supply.

2. The remote monitoring and management rice seedling irrigation and plant protection system according to claim 1, characterized in that: The raw environmental data is preprocessed to generate preprocessed data, specifically in the following steps: A low-pass filter is used to filter and denoise the continuous numerical raw data collected by the soil moisture sensor to remove high-frequency noise interference and obtain filtered soil moisture data. The sliding average filtering algorithm is used to smooth the continuous numerical raw data collected by the soil temperature sensor to eliminate random fluctuations and obtain filtered soil temperature data. The limiting filtering method is used to remove abnormal values ​​from the continuous numerical raw data collected by the soil pH sensor, limiting the data range to a reasonable range, and obtaining the filtered soil pH data; The normalization formula is used to convert the two-dimensional numerical raw data collected by the air temperature and humidity sensor into a standardized format to obtain the standardized air temperature and humidity data; The linear transformation method is used to standardize the continuous numerical raw data collected by the light intensity sensor to obtain the standardized light intensity data; The coordinate transformation formula is used to calculate the horizontal wind speed component of the continuous numerical raw data collected by the three-dimensional wind speed and direction sensor to obtain the standardized wind speed data. The expression is: ; in, is the normalized wind speed data, , are the original data of the horizontal component of wind speed, is the minimum value of the east-west wind speed component in historical records. is the maximum value of the east-west wind speed component in historical records. is the minimum value of the north-south wind speed component in historical records. It is the maximum value of the north-south wind speed component in historical records; The linear scaling method is used to standardize the continuous numerical raw data collected by the tipping bucket rain gauge, and the precipitation data is mapped to the interval [0,1] to obtain the standardized rainfall data. For the discrete numerical raw data collected by agricultural cameras, the pixel intensity ratio calculation formula is directly used for normalization processing to obtain the standardized crop image data; After completing the filtering, denoising and standardization of the above sensor data, a timestamp is added to all the standardized data to generate a preprocessed data set containing time information.

3. The remote monitoring and management rice seedling irrigation and plant protection system according to claim 2, characterized in that: The web and mobile access interfaces are provided to display real-time monitoring data, push decision suggestions, send alarm notifications, record execution logs, and support users to remotely and manually intervene in the execution control module. The specific steps are as follows: Use the front-end development framework React to build the web and mobile application interface UI, and achieve two-way communication with the cloud server through the API interface; The WebSocket protocol is used to achieve real-time data transmission, and the real-time monitoring data from the sensor acquisition module and the execution control module are transmitted to the Web and mobile applications; Use the visual chart library ECharts to graphically display real-time monitoring data and generate intuitive environmental status trend charts and equipment operation status charts; The message push mechanism is used to push the decision suggestions generated by the prediction modeling module to users. The expression is: ； in, A set of decision suggestions, The best irrigation time point is is the water requirement, For fertilizer timing and dosage recommendations, Prioritize plant protection interventions; The alarm rule engine is used to analyze real-time monitoring data. When an abnormal situation is detected, an alarm notification is triggered and sent to the user via SMS or email.

4. The remote monitoring and management rice seedling irrigation and plant protection system according to claim 3, characterized in that: The adaptive learning rate adjustment strategy is used to automatically adjust the learning rate during each iteration to accelerate convergence and improve model accuracy. The specific process is as follows: Get the gradient of the loss function at the current time step, and calculate the loss function based on the training samples of the current batch through the back propagation algorithm Model parameters The gradient of is expressed as: ； in, is a The dimension of the model parameter value at the iteration The same vector, indicating that The partial derivative of the loss function with respect to the parameters at time steps is, represents the model parameters; Calculate the first-order moment estimate of the current gradient, the expression is: ; in, is the first-order moment estimate of the previous time step, Control the retention ratio of historical gradients, is the contribution of the current gradient, Reflects the changing trend of the gradient; Calculate the second-order moment estimate of the current gradient, the expression is: ; in, is the second-order moment estimate of the previous time step, Indicates that each element of the gradient vector is squared. Used to measure the degree of fluctuation of gradient changes, is the second-order moment attenuation coefficient; Since the initial moment and Are all 0, there will be deviations in the first few rounds, and they need to be corrected unbiasedly. The expression is: ; in, is the first-order moment estimate after bias correction, is the bias-corrected second-order moment estimate, and are exponential power terms respectively.

5. The remote monitoring and management rice seedling irrigation and plant protection system according to claim 4, characterized in that: The long short-term memory network LSTM is used as the basic model. During the training of the training sample set, the loss function needs to be calculated, including: During the training process, the back-propagation through time algorithm is used to update the gradient of the LSTM model; The input sequence is divided into time segments of fixed length, which are input into LSTM in sequence. The prediction error of each time step is calculated, and the error is back-propagated to the initial time step through the chain rule. The expression is: ; in, is the total loss function, is the loss function, is the true target value, is the model prediction value, is the total number of time steps, i.e. the length of the sequence.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of remotely monitoring and managing the rice seedling irrigation and plant protection system according to any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of remotely monitoring and managing the rice seedling irrigation and plant protection system according to any one of claims 1 to 5 are implemented.

Images