Intelligent Patrol Inspection System and Method for Agricultural Park Based on Digital Twin

Abstract

The present invention discloses an intelligent inspection system and method for an agricultural park based on digital twin, specifically related to the field of agricultural park management; by using edge computing technology to process sensor data in real time before constructing a three-dimensional model and generating a three-dimensional digital twin model by using a data fusion algorithm, an optimized agricultural management plan and inspection strategy can be generated according to the model prediction. By collecting and analyzing sensor data in real time, the sensors are automatically calibrated to ensure the consistency of the data update frequency and the real-time nature of data synchronization, thereby improving the accuracy of model prediction; according to the accuracy analysis results, the inspection strategy of the park is dynamically adjusted, which can more effectively identify and solve problems in the agricultural park, optimize resource allocation, improve crop yield and quality, and ultimately realize the intelligentization and refinement of agricultural management.

Description

Technical Field

[0001] The present invention relates to the field of agricultural park management, and particularly to an intelligent patrol inspection system and method for agricultural parks based on digital twins. Background Art

[0002] With the development of smart agriculture, digital twin technology has gradually been applied to agricultural production to improve the yield and quality of crops. Digital twin technology creates a virtual model to reflect the state of the physical agricultural park in real time, thus realizing the comprehensive monitoring and management of the agricultural production process. However, when inputting the collected geographical data, crop data, and environmental data from sensors into the digital twin platform to generate a three-dimensional model of the agricultural park, if the data update times of different data sources are inconsistent, it may lead to difficulty in data synchronization, resulting in errors in model prediction. And if there are errors in model prediction, it may not be possible to accurately identify the latest crop problems. At the same time, if crop problems cannot be discovered and handled in a timely manner, it may lead to the deterioration of crop health, and it is impossible to take remedial measures in the early stage, resulting in greater losses. Summary of the Invention

[0003] The purpose of the present invention is to provide an intelligent patrol inspection system and method for agricultural parks based on digital twins to solve the deficiencies in the background art.

[0004] To achieve the above purpose, the present invention provides the following technical solutions: An intelligent patrol inspection method for agricultural parks based on digital twins, including the following steps:

[0005] S1: Before constructing the three-dimensional model, the sensor data is processed in real time through edge computing technology and preliminarily analyzed, and the processed data is input into the digital twin platform;

[0006] S2: The digital twin platform uses a data fusion algorithm to generate a three-dimensional digital twin model, and based on the prediction results of the model, a preliminary agricultural management plan is generated, and the park patrol inspection strategy is determined;

[0007] S3: According to the park patrol inspection strategy, the soil, meteorological, and environmental data in the park are collected in real time through different types of sensors, and the consistency of the update frequencies of different data sources is analyzed, and the sensors are automatically calibrated according to the analysis results;

[0008] S4: The data between the edge computing device and the central data server is integrated, and the real-time synchronization of the integrated data in time and space is analyzed to evaluate whether there is an abnormality in the synchronization of the integrated data;

[0009] S5: The consistency of the update frequencies of different data sources and the real-time synchronization of the integrated data in time and space are comprehensively analyzed to evaluate the accuracy of the prediction results of the three-dimensional digital twin model;

[0010] S6: According to the evaluation results, divide the prediction results of the 3D digital twin model into accurate predictions and inaccurate predictions, and dynamically adjust the park inspection strategy according to the division results.

[0011] Preferably, in S1, the edge computing devices include edge servers, gateway devices, and intelligent routers; the sensors include soil humidity sensors, temperature sensors, light sensors, and meteorological sensors.

[0012] Preferably, the 3D digital twin model is generated by the Bayesian fusion method, specifically:

[0013] Collect real-time sensor observation data, and based on sensor characteristics and observation data, construct a likelihood function, that is, the probability distribution of data under given observation values; apply Bayes' theorem, and calculate the posterior distribution according to the prior distribution and the likelihood function. The Bayes' formula is: where P(θ∣D) is the posterior distribution, P(D∣θ) is the likelihood function, P(θ) is the prior distribution, and P(D) is the evidence; through the posterior distribution, obtain the optimal estimate of the data, generate a comprehensively processed data set; use geographic information system software to perform 3D modeling on the fused data, and use 3D modeling software to generate a 3D model of the agricultural park, reflecting terrain, crop, and environmental information.

[0014] Preferably, in S3, analyze the data update frequencies of each sensor, ensure that the acquisition frequencies of different data sources are consistent, calculate the update frequency of each sensor, compare the difference between it and the preset frequency, analyze the change of the frequency difference of the sensors within a fixed time, and generate an overall update frequency deviation index. The method for obtaining the overall update frequency deviation index is:

[0015] Obtain the data timestamps of each sensor within the s time period, record the time points of each data update, convert the timestamps into time series data, form the data update sequence of each sensor, select a wavelet function to perform discrete wavelet transform on the data update sequence of the sensor, and extract multi-scale frequency components. The transformation expression is: where W j (k) is the wavelet coefficient of the j-th layer, x(n) is the original time series, and ψ j,k (n) is the wavelet function with scale j and translation k; determine the main frequency component through energy density calculation. The specific calculation expression is: E j =∑ k |W j (k)| 2 ; where E j is the wavelet energy of the j-th layer. According to the main frequency component, calculate the actual update frequency of each sensor. In the formula, factual is the actual update frequency, Δd is the time interval corresponding to the main frequency component. Compare the actual update frequency of each sensor with the preset frequency, calculate the frequency deviation, Δfi = |factual(i) - fpreset|; where, Δfi is the frequency deviation of the i-th sensor, factual(i) is the actual update frequency of the i-th sensor, and fpreset is the preset update frequency; calculate the frequency deviations of all sensors, and after summing them up, obtain the overall update frequency deviation index.

[0016] Preferably, in S4, generate a data synchronization real-time index according to the synchronization degree of different data sources in terms of time stamp and geolocation. Then the method for obtaining the data synchronization real-time index is as follows:

[0017] Collect the data time stamp sequences of each sensor to form multiple time sequences T1, T2, …, TN. Organize the data time stamp sequences of each sensor into a unified format to ensure the integrity of the time sequences; for two time sequences Ti = {ti1, ti2, …, tim} and Tj = {tj1, tj2, …, tjq}, construct a distance matrix D, where D[k, h] represents the distance between time stamps tik and tjh, D[k, h] = ∣tik - tjh∣, where tik and tjh are the time stamps in sequences Ti and Tj respectively;

[0018] Construct a cumulative distance matrix C, where C[k, h] represents the minimum cumulative distance from the starting point to position (k, h). The specific calculation expression is: C[k, h] = D[k, h] + min(C[k - 1, h], C[k, h - 1], C[k - 1, h - 1]); where the initial condition is C[1, 1] = D[1, 1]; by backtracking the cumulative distance matrix, find the optimal alignment path W to minimize the total alignment distance between the two time sequences. Among them, W = {(k1, h1), (k2, h2), …, (kp, hp)}; calculate the average alignment distance on the optimal alignment path as the DTW distance between the two time sequences. The specific calculation expression is: p is the total number of paths;

[0019] Perform DTW calculations on all pairs of sensor time sequences to obtain the average DTW distance of all pairs of time sequences. N is the total number of sensors; calculate the difference between the actual position and the expected position of each data point, and calculate the data synchronization real-time index. The specific calculation expression is: DS = αDTW- + βΔs; where α and β are the weight coefficients of time and space synchronization, and the value ranges of α and β are (0, 1); Δs is the difference between the actual position and the expected position of each data point, and DS is the data synchronization real-time index.

[0020] Preferably, compare the obtained data synchronization real-time index with a preset reference threshold of the data synchronization real-time index. If the data synchronization real-time index is greater than or equal to the preset reference threshold of the data synchronization real-time index, the synchronization real-time of different data sources in terms of timestamp and geolocation is high, and the integrated data synchronization is in a normal state. At this time, a synchronization normal signal is generated. If the data synchronization real-time index is less than the preset reference threshold of the data synchronization real-time index, the synchronization real-time of different data sources in terms of timestamp and geolocation is low, and the integrated data synchronization is in an abnormal state. At this time, a synchronization abnormal signal is generated.

[0021] Preferably, in S5, convert the overall update frequency deviation index and the data synchronization real-time index into a first feature vector, and use the first feature vector as the input of a machine learning model. The machine learning model takes predicting the accuracy value label of the prediction result of the three-dimensional digital twin model for each group of first feature vectors as the prediction target, and takes minimizing the sum of the prediction errors of the accuracy value labels of all the prediction results of the three-dimensional digital twin models as the training target. Train the machine learning model until the sum of the prediction errors reaches convergence and then stop the model training. Determine the accuracy value of the prediction result of the three-dimensional digital twin model according to the model output result.

[0022] Preferably, collect the accuracy values of all the prediction results of the three-dimensional digital twin models within a obtained fixed time period, establish a corresponding data combination, calculate the mean and standard deviation of the data set, analyze the mean and standard deviation of the accuracy values of the prediction results of the three-dimensional digital twin models, and divide the accuracy of the model prediction results according to the analysis results, dividing it into accurate prediction and inaccurate prediction.

[0023] Preferably, if the mean of the accuracy values in the data set is greater than or equal to the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is less than the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is high and stable, and it is divided into accurate prediction. At this time, no warning signal is generated, and there is no need to frequently adjust the inspection strategy.

[0024] If the mean of the accuracy values is greater than or equal to the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is greater than or equal to the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is high but unstable, and it is divided into inaccurate prediction. At this time, a third-level warning signal is generated, and the inspection strategy needs to be adjusted to pay attention to the unstable prediction area.

[0025] If the mean of the accuracy values is less than the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is greater than or equal to the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is low and unstable, and it is divided into inaccurate prediction. At this time, a first-level warning signal is generated, and the model needs to be re-evaluated and corrected, and the inspection strategy needs to be significantly adjusted to focus on the inaccurate area.

[0026] If the mean value of the accuracy value is less than the reference threshold of the mean value of the accuracy value, and the standard deviation of the accuracy value is less than the reference threshold of the standard deviation of the accuracy value, the accuracy of the prediction result is low but stable. It is classified as an accuracy prediction. At this time, a secondary warning signal is generated, and the model needs to be adjusted and optimized, and the overall inspection strategy needs to be improved to improve the prediction accuracy.

[0027] The present invention also provides an intelligent inspection system for an agricultural park based on digital twin, including an edge computing module, a data fusion module, an update frequency analysis module, a data synchronization analysis module, an accuracy evaluation module, and an inspection strategy adjustment module;

[0028] Edge computing module: Before constructing the three-dimensional model, the sensor data is processed in real time through edge computing technology and preliminary analysis is carried out, and the processed data is input into the digital twin platform;

[0029] Data fusion module: The digital twin platform uses a data fusion algorithm to generate a three-dimensional digital twin model. According to the prediction result of the model, an agricultural management plan is initially generated, and the inspection strategy for the park is determined;

[0030] Update frequency analysis module: According to the inspection strategy of the park, the soil, meteorological, and environmental data in the park are collected in real time through different types of sensors, and the consistency of the update frequencies of different data sources is analyzed. The sensors are automatically calibrated according to the analysis results;

[0031] Data synchronization analysis module: Integrate the data between the edge computing device and the central data server, analyze the real-time synchronization of the integrated data in time and space, and evaluate whether there is an abnormality in the synchronization of the integrated data;

[0032] Accuracy evaluation module: Comprehensively analyze the consistency of the update frequencies of different data sources and the real-time synchronization of the integrated data in time and space, and evaluate the accuracy of the prediction result of the three-dimensional digital twin model;

[0033] Inspection strategy adjustment module: According to the evaluation results, divide the prediction results of the three-dimensional digital twin model into accuracy predictions and inaccurate predictions, and dynamically adjust the inspection strategy of the park according to the division results.

[0034] In the above technical solution, the technical effects and advantages provided by the present invention:

[0035] 1. The present invention realizes the comprehensive monitoring and refined management of agricultural parks through edge computing technology, data fusion algorithms, and dynamically adjusted inspection strategies. Specifically, the system can process and analyze sensor data in real time, generate a three-dimensional digital twin model, and formulate optimized agricultural management plans based on the prediction results of the model. By ensuring the synchronization and consistency of data sources, the accuracy of model prediction is effectively improved, thereby optimizing resource allocation and enhancing crop yield and quality.

[0036] 2. The present invention dynamically adjusts the inspection strategy by automatically calibrating sensors and comprehensively analyzing the data update frequency and synchronization, enabling the timely identification and handling of abnormal situations in agricultural parks. Such a system not only improves the efficiency and effectiveness of agricultural production but also reduces the need for manual intervention and operating costs. At the same time, the early warning mechanism and prediction result classification further enhance the reliability and practicality of the system, making agricultural management more intelligent and scientific, thus bringing significant economic and social benefits to agricultural production. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0038] Figure 1 It is a flowchart of the method of the present invention.

[0039] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0041] Example 1, please refer to Figure 1 As shown, the intelligent inspection method for agricultural parks based on digital twins in this embodiment includes the following steps:

[0042] S1: Before constructing the three-dimensional model, the sensor data is processed in real time through edge computing technology and preliminarily analyzed, and the processed data is input into the digital twin platform;

[0043] S2: The digital twin platform uses data fusion algorithms to generate a 3D digital twin model. Based on the prediction results of the model, an agricultural management plan is initially generated, and the park inspection strategy is determined.

[0044] S3: According to the park inspection strategy, soil, meteorological, and environmental data in the park are collected in real time through different types of sensors, and the consistency of the update frequencies of different data sources is analyzed. The sensors are automatically calibrated according to the analysis results.

[0045] S4: Integrate the data between the edge computing device and the central data server, analyze the synchronization real-time performance of the integrated data in terms of time and space, and evaluate whether there are any abnormalities in the integrated data synchronization.

[0046] S5: Conduct a comprehensive analysis of the consistency of the update frequencies of different data sources and the synchronization real-time performance of the integrated data in terms of time and space, and evaluate the accuracy of the prediction results of the 3D digital twin model.

[0047] S6: According to the evaluation results, divide the prediction results of the 3D digital twin model into accurate predictions and inaccurate predictions, and dynamically adjust the park inspection strategy according to the division results.

[0048] Among them, in S1, before constructing the 3D model, the sensor data is processed in real time through edge computing technology and preliminary analysis is carried out, and the processed data is input into the digital twin platform.

[0049] According to the specific requirements of the agricultural park, select appropriate edge computing devices, such as edge servers, gateway devices, or intelligent routers. Install the edge computing devices at key locations in the agricultural park to ensure that the signal ranges of all sensors can be covered.

[0050] Deploy various sensors, including soil moisture sensors, temperature sensors, light sensors, meteorological sensors, etc. Set the data collection frequency of the sensors to ensure real-time data acquisition.

[0051] Conduct preliminary data filtering and cleaning on the edge computing device to remove noise and outliers and ensure data quality. Use simple statistical methods (such as mean, median) or rules (such as threshold filtering) to process abnormal data.

[0052] Convert the sensor data into a unified format for convenient subsequent data processing and transmission. Standardize the data structure to ensure the compatibility of data from different sources.

[0053] Utilize the computing power of the edge computing device to process the collected data in real time. Conduct simple data analysis, such as calculating the average value, maximum value, minimum value, etc., to initially understand the data characteristics.

[0054] Perform a preliminary analysis and judgment on the data through the set algorithms and rules. For example, monitor the change trend of soil moisture and determine whether irrigation is needed. If abnormal situations (such as too low soil moisture, abnormal temperature, etc.) are detected, immediately trigger an alarm or corresponding treatment measures.

[0055] Compress the data after real-time processing to reduce the bandwidth requirements for data transmission. Adopt data compression algorithms, such as lossless compression or lossy compression, and select the appropriate method according to the application scenario.

[0056] Through wireless communication technologies (such as LoRa, NB-IoT, Wi-Fi, 5G, etc.), transmit the processed data from the edge computing device to the digital twin platform. Ensure the reliability and real-time nature of data transmission, and avoid data loss and transmission delays.

[0057] S2: The digital twin platform uses data fusion algorithms to generate a three-dimensional digital twin model. According to the prediction results of the model, initially generate an agricultural management plan and determine the park inspection strategy.

[0058] Collect real-time data from sensors such as soil moisture, temperature, meteorology, and light, and perform preprocessing to clean and standardize the data. Obtain high-resolution drone and satellite images to provide a comprehensive view of the geography and crop status.

[0059] Apply data fusion algorithms to comprehensively process the preprocessed sensor data, drone, and satellite images. Commonly used data fusion algorithms include Kalman filtering, Bayesian fusion method, and deep learning methods.

[0060] Generate a three-dimensional digital twin model through the Bayesian fusion method, specifically:

[0061] Define the prior distribution according to historical data and expert knowledge. For example, based on the data of soil moisture, temperature, etc. accumulated over the years, define its prior distribution.

[0062] Collect real-time sensor observation data, and based on the sensor characteristics and observation data, construct a likelihood function to represent the probability distribution of the data under the given observed values.

[0063] Apply Bayes' theorem to calculate the posterior distribution according to the prior distribution and the likelihood function. Bayes' formula: Where P(θ∣D) is the posterior distribution, P(D∣θ) is the likelihood function, P(θ) is the prior distribution, and P(D) is the evidence.

[0064] Through the posterior distribution, the optimal estimate of the data is obtained, and a comprehensively processed dataset is generated. Use Geographic Information System (GIS) software (such as ArcGIS) to perform 3D modeling on the fused data. Use 3D modeling software (such as AutoCAD, 3ds Max) to generate a 3D model of the agricultural park, reflecting terrain, crop, and environmental information.

[0065] According to the prediction results of the model, an optimized agricultural management plan is generated, including irrigation plans, fertilization programs, pest control measures, etc. Irrigation management: Based on soil moisture and weather forecasts, generate accurate irrigation plans to avoid water resource waste. Fertilization management: Analyze soil nutrient content and crop requirements to develop a scientific fertilization plan to improve crop yield and quality. Pest control: Utilize pest prediction results to take preventive measures in advance to reduce the amount of pesticide used and environmental impact.

[0066] Based on the 3D model and prediction results, identify areas that need to be inspected with priority, such as plots with poor crop health, high-risk pest and disease areas, etc. Determine inspection frequency: Based on the crop growth cycle, environmental changes, and prediction results, determine the inspection frequency and develop a detailed inspection plan.

[0067] Use the 3D model to plan the flight path of the drone for inspection to ensure coverage of all key areas. Use path optimization algorithms (such as A* algorithm, Dijkstra algorithm) to generate an efficient inspection route. During the inspection process, monitor the inspection situation of the drone in real time, and dynamically adjust the inspection strategy according to real-time data to ensure timely detection and handling of anomalies. According to the 3D model and inspection requirements, optimize the layout positions of ground sensors to ensure the comprehensiveness and accuracy of data collection. Use automated inspection equipment (such as automated spraying systems, ground robots) for inspection to cooperate with the drone to achieve full-range monitoring.

[0068] S3: According to the park inspection strategy, use different types of sensors to collect soil, meteorological, and environmental data in the park in real time, analyze the consistency of the update frequencies of different data sources, and automatically calibrate the sensors according to the analysis results.

[0069] Based on the 3D digital twin model, determine the areas in the agricultural park that need to be monitored with priority, such as key crop growth areas, high-incidence pest and disease areas, etc. Deploy different types of sensors in the target areas, including soil moisture sensors, meteorological sensors, environmental sensors, etc.

[0070] Sensor types include: Soil sensors: Used to monitor soil moisture, temperature, and nutrient content. Meteorological sensors: Used to monitor meteorological conditions such as temperature, humidity, wind speed, wind direction, rainfall, and solar radiation. Environmental sensors: Used to monitor environmental factors such as light intensity, carbon dioxide concentration, and air quality.

[0071] Analyze the data update frequencies of each sensor to ensure that the acquisition frequencies of different data sources are consistent. Calculate the update frequency of each sensor and compare the difference between it and the preset frequency. Analyze the change of the frequency difference of the sensor within a fixed time and generate an overall update frequency deviation index. The method for obtaining the overall update frequency deviation index is as follows:

[0072] Obtain the data timestamps of each sensor within the s time period, record the time points of each data update, convert the timestamps into time series data, and form a data update sequence for each sensor. Select an appropriate wavelet function (such as Haar wavelet, Daubechies wavelet) to perform discrete wavelet transform on the data update sequence of the sensor, extract the frequency components at multiple scales, and the transformation expression is: Where, W j (k) is the wavelet coefficient of the j-th layer, x(n) is the original time series, and ψ j,k (n) is the wavelet function with scale j and translation k; by analyzing the wavelet coefficients, determine the main frequency component, and determine the main frequency component through energy density calculation. The specific calculation expression is: E j =∑ k |W j (k)| 2 ; where, E j is the wavelet energy of the j-th layer. According to the main frequency component, calculate the actual update frequency of each sensor, In the formula, factual is the actual update frequency, Δd is the time interval corresponding to the main frequency component. Compare the actual update frequency of each sensor with the preset frequency and calculate the frequency deviation, Δfi = |factual(i)-fpreset|; where, Δfi is the frequency deviation of the i-th sensor, factual(i) is the actual update frequency of the i-th sensor, and fpreset is the preset update frequency; calculate the frequency deviations of all sensors, and after summing them up, obtain the overall update frequency deviation index.

[0073] The larger the overall update frequency deviation index, the worse the consistency of the update frequencies of different data sources. This means that there are large differences in the time intervals of data acquisition by each sensor, resulting in difficulties in time alignment of data and affecting the effect of data fusion. Sensors with large frequency deviations have inconsistent updates, which will lead to instability of the overall data, and thus affect the accuracy and real-time performance of the digital twin model.

[0074] Frequency inconsistency may be caused by various factors, including sensor hardware failures, environmental interference, or communication delays, etc. These factors will cause the sensor to be unable to collect data at the preset frequency, resulting in deviations. The larger the deviation index, the more serious these problems are and the greater the difficulty of data synchronization.

[0075] To improve the consistency of the data source update frequency, it is necessary to calibrate and maintain the sensors regularly and adopt advanced algorithms (such as adaptive sliding window or Bayesian update) to dynamically adjust and optimize the data acquisition frequency. This will help reduce the overall update frequency deviation index and ensure the accuracy of data fusion and model prediction.

[0076] S4: Integrate the data between the edge computing device and the central data server, analyze the synchronization real-time performance of the integrated data in terms of time and space, and evaluate whether there are any anomalies in the integrated data synchronization.

[0077] Integrating the data between the edge computing device and the central data server means aggregating and transmitting the data scattered on the edge computing device to the central data server. This process includes standardizing the data from different sources and in different formats to ensure the consistency and integrity of the data during transmission. Real-time data transmission is achieved through wireless communication technologies (such as LoRa, NB-IoT, Wi-Fi, 5G, etc.), enabling the data to be timely converged to the central server for further analysis and processing.

[0078] Analyzing the synchronization real-time performance of the integrated data in terms of time and space refers to evaluating the synchronization real-time performance of the data in terms of time and geographical location. The synchronization check in terms of time ensures that the timestamps of different data sources are consistent, solving the data deviation caused by time asynchronization; the synchronization check in terms of space ensures the accurate positioning of the data in the geographical space, ensuring that the data from different sources can correctly correspond to the actual geographical locations. Through these analyses, potential problems in data synchronization can be identified and corrected, improving the accuracy and real-time performance of the data, thereby enhancing the reliability and effectiveness of the digital twin model.

[0079] Generate a data synchronization real-time index according to the synchronization degree of different data sources in terms of timestamp and geographical positioning. The method for obtaining the data synchronization real-time index is as follows:

[0080] Collect the data timestamp sequences of each sensor to form multiple time series T1, T2, …, TN. Organize the data timestamp sequences of each sensor into a unified format to ensure the integrity of the time series. For two time series Ti = {ti1, ti2, …, tim} and Tj = {tj1, tj2, …, tjn}, construct a distance matrix D, where D[k, h] represents the distance between timestamps tik and tjh, and D[k, h] = ∣tik - tjh∣, where tik and tjh are the timestamps in sequences Ti and Tj respectively.

[0081] Construct a cumulative distance matrix C, where C[k, h] represents the minimum cumulative distance from the starting point to the position (k, h). The specific calculation expression is: C[k, h] = D[k, h] + min(C[k - 1, h], C[k, h - 1], C[k - 1, h - 1]); where the initial condition is C[1, 1] = D[1, 1]; by backtracking the cumulative distance matrix, find the optimal alignment path W to minimize the total alignment distance of the two time series. Here, W = {(k1, h1), (k2, h2),..., (kp, hp)}; calculate the average alignment distance on the optimal alignment path as the DTW distance between the two time series. The specific calculation expression is: p is the total number of paths;

[0082] Perform DTW calculations on all pairs of sensor time series to obtain the average DTW distance of all time series pairs. N is the total number of sensors; calculate the difference between the actual position and the expected position of each data point, and calculate the data synchronization real-time index. The specific calculation expression is: DS = αDTW- + βΔs; where α and β are the weight coefficients of time and space synchronization, and the value ranges of α and β are (0, 1); Δs is the difference between the actual position and the expected position of each data point, and DS is the data synchronization real-time index.

[0083] Compare the obtained data synchronization real-time index with the pre-set data synchronization real-time index reference threshold. If the data synchronization real-time index is greater than or equal to the pre-set data synchronization real-time index reference threshold, it indicates that the synchronization real-time of different data sources in terms of timestamps and geolocation is high, and the integrated data synchronization is in a normal state. At this time, a synchronization normal signal is generated; if the data synchronization real-time index is less than the pre-set data synchronization real-time index reference threshold, it indicates that the synchronization real-time of different data sources in terms of timestamps and geolocation is low, and the integrated data synchronization is in an abnormal state. At this time, a synchronization abnormal signal is generated.

[0084] S5: Comprehensively analyze the consistency of the update frequencies of different data sources and the synchronization real-time of the integrated data in terms of time and space to evaluate the accuracy of the prediction results of the three-dimensional digital twin model.

[0085] Convert the overall update frequency deviation index and the data synchronization real-time index into a first feature vector, and use the first feature vector as the input of the machine learning model. The machine learning model takes predicting the accuracy value label of the prediction result of the three-dimensional digital twin model for each group of the first feature vectors as the prediction target, and takes minimizing the sum of the prediction errors of the accuracy value labels of all the prediction results of the three-dimensional digital twin model as the training target. Train the machine learning model until the sum of the prediction errors reaches convergence and then stop the model training. Determine the accuracy value of the prediction result of the three-dimensional digital twin model according to the model output result.

[0086] The method for obtaining the accuracy value of the prediction result of the three-dimensional digital twin model is as follows: from the first feature vector training data of the trained machine learning model, obtain the corresponding function expression: MT = f1(KP, DS); where f1 is the output function of the model, KP is the overall update frequency deviation index, DS is the data synchronization real-time index, and MT is the accuracy value of the prediction result of the three-dimensional digital twin model.

[0087] S6: According to the evaluation results, divide the prediction results of the three-dimensional digital twin model into accurate predictions and inaccurate predictions, and dynamically adjust the park patrol strategy according to the division results.

[0088] Collect the accuracy values of all prediction results of the three-dimensional digital twin model within the obtained fixed time period, establish a corresponding data combination, calculate the mean and standard deviation of the data set, analyze the mean and standard deviation of the accuracy values of the prediction results of the three-dimensional digital twin model, and divide the accuracy of the model prediction results according to the analysis results into accurate predictions and inaccurate predictions.

[0089] If the mean of the accuracy values in the data set is greater than or equal to the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is less than the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is high and stable, and it is divided into an accurate prediction. At this time, no warning signal is generated, and there is no need to frequently adjust the patrol strategy;

[0090] If the mean of the accuracy values is greater than or equal to the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is greater than or equal to the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is high but unstable, and it is divided into an inaccurate prediction. At this time, a level-three warning signal is generated, and the patrol strategy needs to be adjusted to pay attention to the unstable prediction area;

[0091] If the mean of the accuracy values is less than the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is greater than or equal to the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is low and unstable, and it is divided into an inaccurate prediction. At this time, a level-one warning signal is generated, and the model needs to be re-evaluated and corrected, and the patrol strategy needs to be significantly adjusted to focus on the inaccurate area;

[0092] If the mean of the accuracy values is less than the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is less than the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is low but stable, and it is divided into an accurate prediction. At this time, a level-two warning signal is generated, and the model needs to be adjusted and optimized, and the overall patrol strategy needs to be improved to improve the prediction accuracy.

[0093] Here, it should be noted that the management level of the first-level warning signal is higher than that of the second-level warning signal, and the management level of the second-level warning signal is higher than that of the third-level warning signal. Relevant personnel can dynamically adjust the inspection strategy of the agricultural park according to the level of the warning signal.

[0094] According to the division results, dynamically adjust the inspection strategy of the agricultural park: for the prediction results with high accuracy and low fluctuation, maintain the existing inspection frequency and key areas.

[0095] For the prediction results with high accuracy but large fluctuations, increase the inspection frequency and focus on the areas with larger fluctuations.

[0096] For the prediction results with low accuracy and high fluctuations, recalibrate and optimize the model, and significantly increase the inspection frequency to cover more areas.

[0097] For the prediction results with low accuracy but small fluctuations, optimize the model and at the same time adjust the inspection strategy to improve the prediction accuracy.

[0098] In this embodiment, before constructing the three-dimensional model, the sensor data is processed in real time through edge computing technology and preliminary analysis is carried out. The processed data is input into the digital twin platform. The digital twin platform uses the data fusion algorithm to generate a three-dimensional digital twin model. According to the prediction results of the model, an agricultural management plan is initially generated and the inspection strategy of the park is determined. According to the inspection strategy of the park, the soil, meteorological and environmental data in the park are collected in real time through different types of sensors, and the consistency of the update frequencies of different data sources is analyzed. The sensors are automatically calibrated according to the analysis results. The data between the edge computing device and the central data server is integrated, and the synchronization real-time performance of the integrated data in time and space is analyzed to evaluate whether there is an abnormality in the synchronization of the integrated data. The consistency of the update frequencies of different data sources and the synchronization real-time performance of the integrated data in time and space are comprehensively analyzed to evaluate the accuracy of the prediction results of the three-dimensional digital twin model. According to the evaluation results, the prediction results of the three-dimensional digital twin model are divided into accurate predictions and inaccurate predictions, and the inspection strategy of the park is dynamically adjusted according to the division results. It can more effectively identify and solve problems in the agricultural park, optimize resource allocation, improve crop yield and quality, and ultimately realize the intelligentization and refinement of agricultural management.

[0099] Embodiment 2, please refer to Figure 2 As shown, the intelligent inspection system of the agricultural park based on digital twin in this embodiment includes an edge computing module, a data fusion module, an update frequency analysis module, a data synchronization analysis module, an accuracy evaluation module, and an inspection strategy adjustment module;

[0100] Edge computing module: Before constructing the three-dimensional model, the sensor data is processed in real time through edge computing technology and preliminary analysis is carried out, and the processed data is input into the digital twin platform;

[0101] Data fusion module: The digital twin platform uses data fusion algorithms to generate a 3D digital twin model. Based on the prediction results of the model, an agricultural management plan is initially generated, and the park inspection strategy is determined.

[0102] Update frequency analysis module: According to the park inspection strategy, soil, meteorological, and environmental data in the park are collected in real time through different types of sensors, and the consistency of the update frequencies of different data sources is analyzed. The sensors are automatically calibrated according to the analysis results.

[0103] Data synchronization analysis module: Integrate data between edge computing devices and the central data server, analyze the real-time synchronization of the integrated data in terms of time and space, and evaluate whether there are any abnormalities in the synchronization of the integrated data.

[0104] Accuracy evaluation module: Comprehensively analyze the consistency of the update frequencies of different data sources and the real-time synchronization of the integrated data in terms of time and space, and evaluate the accuracy of the prediction results of the 3D digital twin model.

[0105] Inspection strategy adjustment module: According to the evaluation results, divide the prediction results of the 3D digital twin model into accurate predictions and inaccurate predictions, and dynamically adjust the park inspection strategy according to the division results.

[0106] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula that is closest to the actual situation. The preset parameters in the formulas are set by technicians in the field according to the actual situation.

[0107] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0108] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context.

[0109] It should be understood that in various embodiments of the present application, the sequence numbers of the above processes do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0110] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this document can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0111] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application.

Claims

1. An intelligent inspection method for agricultural parks based on digital twins, characterized in that: Including the following steps; S1: Before constructing the 3D model, the sensor data is processed in real time through edge computing technology and preliminary analysis is carried out, and the processed data is input into the digital twin platform; S2: The digital twin platform uses a data fusion algorithm to generate a 3D digital twin model. According to the prediction results of the 3D digital twin model, an agricultural management plan is initially generated, and the park inspection strategy is determined; Generate a 3D digital twin model through the Bayesian fusion method, specifically: collect real-time sensor observation data, construct a likelihood function based on sensor characteristics and observation data, that is, the probability distribution of data given the observed values; apply Bayes' theorem to calculate the posterior distribution according to the prior distribution and the likelihood function. The Bayes' formula is: ; where is the posterior distribution, is the likelihood function, is the prior distribution, is the evidence; through the posterior distribution, obtain the optimal estimate of the data and generate a comprehensively processed dataset; use geographic information system software to perform 3D modeling on the fused data, and use 3D modeling software to generate a 3D model of the agricultural park, reflecting terrain, crop, and environmental information; S3: According to the park inspection strategy, the soil, meteorological and environmental data in the park are collected in real time through different types of sensors, and the consistency of the update frequencies of different data sources is analyzed. According to the analysis results, the sensors are automatically calibrated. Specifically: Analyze the data update frequencies of each sensor to ensure that the acquisition frequencies of different data sources are consistent. Calculate the update frequency of each sensor and compare the difference between it and the preset frequency. Analyze the change situation of the frequency difference of the sensors within a fixed time, and generate an overall update frequency deviation index. The method for obtaining the overall update frequency deviation index is: Obtain the data timestamps of each sensor within the s time period, record the time points of each data update, convert the timestamps into time series data, form the data update sequence of each sensor, select a wavelet function to perform discrete wavelet transform on the data update sequence of the sensor, extract multi-scale frequency components, and the transformation expression is: ; where is the wavelet coefficient of the e-th layer, x(n) is the original time series, is the wavelet function with scale e and translation h; Determine the main frequency component through energy density calculation, and the specific calculation expression is: ; where is the wavelet energy of the e-th layer. According to the main frequency component, calculate the actual update frequency of each sensor, ; in the formula, factual is the actual update frequency, is the time interval corresponding to the main frequency component. Compare the actual update frequency of each sensor with the preset frequency, and calculate the frequency deviation, ; where is the frequency deviation of the i-th sensor, is the actual update frequency of the i-th sensor, and fpreset is the preset update frequency; Calculate the frequency deviations of all sensors, and obtain the overall update frequency deviation index after summing them up. S4: Integrate the data between the edge computing device and the central data server, analyze the synchronization real-time performance of the integrated data in terms of time and space, and evaluate whether there is an abnormality in the integration data synchronization. Specifically: Generate a data synchronization real-time index based on the synchronization degree of different data sources in terms of timestamp and geolocation. The method for obtaining the data synchronization real-time index is as follows: Collect the data timestamp sequences of each sensor to form multiple time series T1, T2, …, TN. Organize the data timestamp sequences of each sensor into a unified format to ensure the integrity of the time series. For two time series and , construct a distance matrix D, where D[k, h] represents the distance between timestamps tik and tjh, where tik and tjh are the timestamps in time series Ti and Tj respectively; construct a cumulative distance matrix C, where C[k, h] represents the minimum cumulative distance from the starting point to position (k, h). The specific calculation expression is: C[k, h]=D[k, h]+min(C[k−1, h], C[k, h−1], C[k−1, h−1]); where the initial condition is C[1, 1]=D[1, 1]; By backtracking the cumulative distance matrix, find the optimal alignment path W to minimize the total alignment distance between the two time series, where W={(k1, h1), (k2, h2), …, (kp, hp)}; Calculate the average alignment distance on the optimal alignment path as the distance between the two time series , and the specific calculation expression is: ; p is the total number of paths; Perform DTW calculations on all pairs of data timestamp sequences of the sensors to obtain the average DTW distance of all time series pairs, denoted as , ; N is the total number of sensors; Calculate the difference between the actual position and the expected position of each data point, and calculate the data synchronization real-time index. The specific calculation expression is: ; where α and β are the weight coefficients of time and space synchronization, and the value ranges of α and β are (0, 1); is the difference between the actual position and the expected position of each data point, and DS is the data synchronization real-time index; S5: Conduct a comprehensive analysis of the consistency of the update frequencies of different data sources and the synchronization real-time performance of the integrated data in terms of time and space, and evaluate the accuracy of the prediction results of the 3D digital twin model; Convert the overall update frequency deviation index and the data synchronization real-time index into a first feature vector, and use the first feature vector as the input of the machine learning model. The machine learning model takes predicting the accuracy value label of the prediction results of the 3D digital twin model for each group of the first feature vectors as the prediction target, and takes minimizing the sum of the prediction errors of the accuracy value labels of all the prediction results of the 3D digital twin model as the training target, and trains the machine learning model until the sum of the prediction errors reaches convergence and then stops the training of the machine learning model. Determine the accuracy value of the prediction results of the 3D digital twin model according to the output result of the machine learning model; S6: According to the evaluation results, divide the prediction results of the 3D digital twin model into accurate predictions and inaccurate predictions, and dynamically adjust the park inspection strategy according to the division results.

2. The intelligent inspection method for agricultural parks based on digital twins according to claim 1, wherein: In S1, the edge computing device includes an edge server, a gateway device, and an intelligent router; the sensors include a soil moisture sensor, a temperature sensor, a light sensor, and a meteorological sensor.

3. The intelligent inspection method for agricultural parks based on digital twins according to claim 1, wherein: Compare the obtained data synchronization real-time index with the pre-set data synchronization real-time index reference threshold. If the data synchronization real-time index is greater than or equal to the pre-set data synchronization real-time index reference threshold, the synchronization real-time performance of different data sources in terms of timestamp and geolocation is high, and the integration data synchronization is in a normal state. At this time, a synchronization normal signal is generated; if the data synchronization real-time index is less than the pre-set data synchronization real-time index reference threshold, the synchronization real-time performance of different data sources in terms of timestamp and geolocation is low, and the integration data synchronization is in an abnormal state. At this time, a synchronization abnormal signal is generated.

4. The intelligent inspection method for agricultural parks based on digital twins according to claim 1, wherein: Collect the accuracy values of all three-dimensional digital twin model prediction results within a fixed time period, establish a corresponding data combination, calculate the mean and standard deviation of the data set, analyze the mean and standard deviation of the accuracy values of the three-dimensional digital twin model prediction results, and divide the accuracy of the three-dimensional digital twin model prediction results according to the analysis results, dividing it into accurate prediction and inaccurate prediction.

5. The intelligent inspection method for agricultural parks based on digital twins according to claim 4, characterized in that: If the mean of the accuracy values in the data set is greater than or equal to the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is less than the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is high and stable, and it is divided into accurate prediction. At this time, no warning signal is generated, and there is no need to frequently adjust the inspection strategy. If the mean of the accuracy values is greater than or equal to the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is greater than or equal to the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is high but unstable, and it is divided into inaccurate prediction. At this time, a level-three warning signal is generated, and the inspection strategy needs to be adjusted to pay attention to the unstable prediction area. If the mean of the accuracy values is less than the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is greater than or equal to the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is low and unstable, and it is divided into inaccurate prediction. At this time, a level-one warning signal is generated, and the three-dimensional digital twin model needs to be re-evaluated and corrected, and the inspection strategy needs to be significantly adjusted to focus on inaccurate areas. If the mean of the accuracy values is less than the reference threshold of the mean of the accuracy values, and the standard deviation of the accuracy values is less than the reference threshold of the standard deviation of the accuracy values, the accuracy of the prediction result is low but stable, and it is divided into accurate prediction. At this time, a level-two warning signal is generated, and the three-dimensional digital twin model needs to be adjusted and optimized, and the overall inspection strategy needs to be improved to enhance the prediction accuracy.

6. The intelligent inspection system for agricultural parks based on digital twins is used to implement the intelligent inspection method for agricultural parks based on digital twins according to any one of claims 1-5, and is characterized in that: It includes an edge computing module, a data fusion module, an update frequency analysis module, a data synchronization analysis module, an accuracy evaluation module, and an inspection strategy adjustment module. Edge computing module: Before constructing the three-dimensional model, the sensor data is processed in real time through edge computing technology and preliminary analysis, and the processed data is input into the digital twin platform. Data fusion module: The digital twin platform uses a data fusion algorithm to generate a three-dimensional digital twin model, and based on the prediction results of the three-dimensional digital twin model, a preliminary agricultural management plan is generated, and the inspection strategy for the park is determined. Update frequency analysis module: According to the park inspection strategy, the soil, meteorological, and environmental data in the park are collected in real time through different types of sensors, and the consistency of the update frequencies of different data sources is analyzed. The sensors are automatically calibrated according to the analysis results. Data synchronization analysis module: Integrate the data between the edge computing device and the central data server, analyze the synchronization real-time performance of the integrated data in terms of time and space, and evaluate whether there are any abnormalities in the integrated data synchronization. Accuracy evaluation module: Comprehensively analyze the consistency of the update frequencies of different data sources and the synchronization real-time performance of the integrated data in terms of time and space to evaluate the accuracy of the prediction results of the three-dimensional digital twin model. Patrol inspection strategy adjustment module: According to the evaluation results, divide the prediction results of the three-dimensional digital twin model into accurate predictions and inaccurate predictions, and dynamically adjust the park patrol inspection strategy according to the division results.

Images