Agricultural and sideline products project management method based on big data technology

Abstract

This invention discloses a project management method for agricultural and sideline products based on big data technology, belonging to the field of product data management. This method collects multi-dimensional spatiotemporal natural attribute data of agricultural and sideline product raw materials, generates a flexible bill of materials with attribute fluctuation ranges through Gaussian mixture model clustering, and realizes the structured expression of non-standard materials. Using the flexible bill of materials as state input, it outputs process route adjustment actions through a reinforcement learning strategy network, realizing dynamic reconstruction of cross-entity process routes. It constructs a temporal knowledge graph to perform entity alignment and credibility scoring on data related to commerce, logistics, information flow, business flow, and capital flow. Data meeting the conditions is mapped to project milestone nodes in the product data management system, achieving automatic progress updates. This invention can be used for product data management and process management of agricultural and sideline product processing projects.

Description

Technical Field

[0001] This invention relates to the field of big data processing technology, specifically to a method for managing agricultural and sideline product projects based on big data technology. It belongs to the cross-technical field of product data management and smart agricultural supply chain management, and particularly to a method for digitally managing the bill of materials, process routes, and project progress milestones in agricultural and sideline product processing projects. It can be applied to scenarios of collaborative management and control between product data management systems and agricultural and sideline product supply chains. Background Technology

[0002] With the deep integration of the digital economy and agricultural modernization, the agricultural and sideline product supply chain is gradually developing towards standardization and project-based management. Against this backdrop, product data management systems, originally applied to the industrial manufacturing sector, have been introduced into the management of agricultural and sideline product processing projects. These systems are used to uniformly manage the entire lifecycle of agricultural and sideline products, from raw material acquisition to finished product delivery. In the industrial manufacturing sector, product data management systems have formed a relatively mature technical system. By designing a hierarchical and structured bill of materials (BOM) model consisting of a bill of materials (BOM), a process BOM, and a manufacturing BOM, standardized parts are uniformly coded and managed. Relying on fixed process route templates and scheduling tools, the system provides visualized management of production project task nodes, resource allocation, and progress plans. In recent years, infrastructure such as agricultural IoT technology, multispectral detection sensors, smart weighbridges, and data interfaces from local weather stations have gradually become widespread, providing a hardware foundation for the digital collection of raw material attributes of agricultural and sideline products.

[0003] A similar existing technical solution is to directly transplant a commercial product data management platform from the industrial field to the management of agricultural and sideline product processing projects. This solution usually involves manually creating standard bill of materials master data entries for raw materials according to fixed material coding rules, setting fixed process route templates for each type of agricultural and sideline product and solidifying them into the project plan. Each node in the supply chain manually enters data such as progress, inventory, and logistics status into the system, and the project management personnel summarize, verify, and manually update the project progress nodes.

[0004] The aforementioned existing technical solutions have the following shortcomings: First, traditional product data management systems are built on a deterministic bill of materials (BOM) formed by design, process, and manufacturing. However, agricultural and sideline products, as materials with non-standardized natural attributes, exhibit natural fluctuations in their physicochemical indicators such as moisture content and maturity depending on the place of origin, climate, and batch. Therefore, the fixed hierarchical coding method used for industrial products cannot be applied to BOM modeling. Due to the structural contradiction between the continuous fluctuation of raw material attributes and the discrete rigidity of fixed coding rules, a large amount of non-standard attribute data cannot be compatible with the product data management system. Second, the processing routes of agricultural and sideline product raw materials exhibit high-frequency dynamic changes depending on the batch characteristics of the raw materials. For example, the same type of fruits and vegetables... Different cleaning, drying, and slicing processes are required due to variations in moisture content and hardness. The fixed process route templates and project schedules in the existing product data management system are difficult to adapt to the dynamic adjustment requirements once set. The contradiction between the static preset characteristics of the process route templates and the dynamic differences in the characteristics of raw material batches makes it difficult to coordinate production across entities and causes deviations in project progress tracking. Thirdly, the level of digitalization varies among the nodes of the agricultural and sideline product supply chain. Source data mainly relies on manual entry. Due to the lack of a unified automatic fusion and trust assessment mechanism for heterogeneous data from multiple entities, data silos and information distortion are easily caused. The progress, cost, and quality data presented by the project management system deviate from the actual business flow of the supply chain. Summary of the Invention

[0005] The technical problem this invention aims to solve is that the rigid bill of materials (BOM) model and fixed process route template used in existing product data management systems cannot adapt to the material structure fluctuations and dynamic changes in processing technology caused by the non-standardized natural properties of agricultural and sideline product raw materials. Furthermore, data among multiple entities in the agricultural and sideline product supply chain is difficult to automatically align, resulting in problems such as the inability to express the BOM in a structured manner, difficulties in cross-entity collaborative production scheduling, and distortion of project progress tracking in agricultural and sideline product project management.

[0006] To address the aforementioned technical problems, this invention provides a method for managing agricultural and sideline product projects based on big data technology, comprising the following:

[0007] The system collects multidimensional spatiotemporal natural attribute raw data of agricultural and sideline products upon arrival. This raw data is preprocessed to obtain a standardized dataset. Feature vectors are constructed by batch, and a Gaussian mixture model is used to cluster the feature vectors of multiple batches to obtain multiple clusters. Attribute fluctuation ranges are generated based on the mean and standard deviation of samples within each attribute dimension, serving as the corresponding fuzzy master data level for that cluster. The feature vectors of the batches to be processed are assigned to the cluster with the highest posterior probability. Flexible bill of materials (BOM) entries are generated based on the corresponding fuzzy master data level. Version numbers are assigned to these flexible BOM entries and bound to batch traceability codes, collection timestamps, and raw attribute data. The BOM entries are then pushed to the material master data module of the product data management system, achieving structured representation and version traceability management of non-standard attribute materials in agricultural and sideline products.

[0008] Using the attribute fluctuation range of flexible bill of materials items, real-time capacity data of each participant in the supply chain, and logistics delay risk factors as state inputs, a reinforcement learning state vector is constructed. The policy network outputs process route adjustment actions based on the state vector. The adjustment actions include adding and deleting process nodes, adjusting the time allocation of process nodes, and reallocating processing entities. A reward function is constructed based on the total project duration, resource utilization rate, and quality pass rate after the process route is executed. The network parameters of the policy network are iteratively updated, and the updated process route output by the policy network is pushed to the project planning module of the product data management system, realizing dynamic and elastic reconstruction of cross-entity process routes.

[0009] The system extracts raw event data related to the current project from the commercial flow system, logistics system, information flow system, business flow system, and capital flow system. It performs entity identification on the raw event data, aligns cross-system entities pointing to the same batch and the same project node, and marks them with timestamps to form a time-series knowledge graph. It calculates the credibility score for data records of the same entity appearing in multiple data sources, and maps data records with credibility scores that meet the preset threshold conditions to the corresponding milestone nodes in the project plan of the product data management system according to the time sequence relationship and entity alignment results. It updates the progress status, completion time, and associated cost and quality data of the milestone node, so as to achieve automatic and accurate alignment between project management progress and underlying business flow data.

[0010] This invention constructs a flexible bill of materials dynamic instantiation mechanism, transforming the continuously fluctuating natural attributes of agricultural and sideline products into a fuzzy master data hierarchy with limited ranges. This enables non-standard agricultural and sideline raw materials to be expressed and version traced in a structured manner in the product data management system, overcoming the defect that rigid bill of materials models cannot be compatible with non-standard attribute data. Through a reinforcement learning-driven cross-entity process route dynamic reconstruction mechanism, the process route and time allocation can be flexibly adjusted according to the raw material batch attributes and the real-time capacity of each entity, overcoming the defect that fixed process route templates cannot adapt to dynamic processing needs, and improving the efficiency of cross-entity collaborative production scheduling. By integrating a five-flow data entity alignment and credibility scoring mechanism based on temporal knowledge graphs, the automatic fusion and filtering of heterogeneous data from multiple entities in the supply chain is realized, replacing manual reporting to complete the automatic updating of project milestones, and improving the authenticity and timeliness of project progress, cost, and quality data.

[0011] Meanwhile, the attribute fluctuation range of the flexible bill of materials serves as the input to the reinforcement learning state vector, directly driving the process route adjustment. The milestone credibility score data output by the knowledge graph is transmitted back to the flexible bill of materials version entry stage and the process route execution feedback stage via a standard interface, forming a closed-loop technical effect of mutual feedback among material attribute identification, process route adjustment, and progress data verification. This overcomes the technical bias of existing technologies that allow for independent and static management of bill of materials, process routes, and progress data, providing a technical foundation for the deep integration of agricultural and sideline product project management and product data management systems. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating the dynamic instantiation process of the flexible bill of materials according to the present invention.

[0013] Figure 2 This is a flowchart illustrating the dynamic reconfiguration of the cross-subject process route in this invention.

[0014] Figure 3 This is a flowchart of the automatic alignment process for the five-stream integration milestones in this invention. Detailed Implementation

[0015] The following is in conjunction with the appendix Figure 1-3 The specific embodiments of the present invention will be further described below. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0016] Example 1: System Overall Architecture

[0017] The overall system upon which the method described in this invention relies consists of a data acquisition layer, a big data platform layer, a standard interface layer, and a product data management application layer. The data acquisition layer comprises multispectral detection equipment, smart weighbridges, origin weather station interfaces, IoT sensors, and the business systems of various participants in the supply chain. It is responsible for collecting data on the natural attributes of raw materials, production capacity, logistics, and business flow. The big data platform layer includes a data cleaning and fusion submodule, a flexible bill of materials generation engine, a process route reinforcement learning engine, and a time-series knowledge graph construction and confidence scoring engine. These engines collaborate through data sharing and feedback mechanisms. The standard interface layer parses the flexible bill of materials, dynamic process routes, and milestone alignment results output by the platform into a standardized data format and encapsulates them through application programming interfaces (APIs). The product data management application layer is an existing commercial product data management system. After receiving data pushed by the standard interface, it automatically triggers project plan updates, bill of materials version entry, and schedule redrawing. The above four-layer architecture interacts with data through a unified big data platform, enabling the collaborative operation of the three core mechanisms of this invention.

[0018] Example 2: Dynamic Instantiation Mechanism for Flexible Bill of Materials

[0019] This mechanism solves the problem that the non-standard attributes of agricultural and sideline products cannot be structurally expressed in the product data management system. Its core is to use big data clustering algorithms to transform continuously fluctuating natural attributes into fuzzy master data levels with limited intervals.

[0020] In the process of receiving agricultural and sideline product raw materials, the raw data of multidimensional spatiotemporal natural attributes, including soil indicators, historical climate data, maturity and moisture content, are collected by using equipment such as near-infrared spectrometers, smart weighbridges and meteorological station interfaces in the production area. Missing values ​​are imputed and outliers are removed to form a standardized raw dataset. The preprocessed multidimensional attribute data is organized into feature vectors by batch, and the feature vector of the i-th batch is denoted as a vector in d-dimensional real space.

[0021] Using the set of feature vectors as input, a Gaussian mixture model is constructed to fit the data distribution, and its probability density function is expressed as:

[0022] ;

[0023] in, This represents the feature vector to be fitted. This represents the total number of Gaussian components, i.e., the number of candidate clusters. Indicates the first The mixture weights of the Gaussian components are such that the sum of all the mixture weights is one. The mean is The covariance matrix is The multidimensional Gaussian probability density function, Indicates the first The central values ​​of each cluster along each attribute dimension Indicates the first The attribute fluctuation covariance matrix of each cluster.

[0024] To determine the number of candidate clusters, the Bayesian information criterion is used to evaluate the model under different numbers of candidate clusters. The evaluation index is expressed as:

[0025] ;

[0026] in, Indicates the number of clusters. The Bayesian information criterion score at that time. Indicates the number of clusters. The maximum likelihood function value of the model under the given conditions Indicates the number of clusters. The number of free parameters in the time model, This represents the total number of batch samples participating in clustering. The number of candidate clusters whose evaluation index values ​​meet the preset optimization conditions is selected as the cluster number. This iterative evaluation process continues until the preset conditions are met and then ends.

[0027] For each cluster, attribute fluctuation ranges are generated based on the mean and standard deviation of the samples within that cluster across each attribute dimension:

[0028] ;

[0029] in, Indicates the first The mean of each cluster along its corresponding attribute dimension. Indicates the first The standard deviation of each cluster along its corresponding attribute dimension This is a configurable interval width coefficient used to adjust the width of the attribute fluctuation interval, which serves as the fuzzy master data level corresponding to the cluster.

[0030] The feature vector of the current batch of goods is assigned to a cluster whose posterior probability meets the preset matching conditions. Based on the fuzzy master data interval of the cluster, a flexible bill of materials (BOM) entry for this batch is generated. A version number is assigned to the flexible BOM entry, and the version number is bound to the batch traceability code, collection timestamp, and original attribute data and stored in the database to achieve structured representation and version traceability. The generated flexible BOM entry is parsed into a standardized data format and pushed to the material master data module of the product data management system through a standard interface, triggering the automatic update of the BOM tree in the product data management system.

[0031] In another implementation, the flexible bill of materials generation process can also be implemented using a density clustering algorithm. By setting neighborhood radius and sample density threshold parameters, the attribute data of agricultural and sideline products with non-elliptical distribution or noise points can be clustered. This can also achieve adaptive generation of fuzzy master data intervals, serving as a supplementary implementation method to the Gaussian mixture model clustering approach.

[0032] Example 3: Dynamic Reconstruction of Cross-Agent Process Routes Based on Reinforcement Learning

[0033] This mechanism addresses the issue of fixed process routes in cross-entity collaborative production scheduling. Its core is to use a flexible bill of materials as the state input and reinforcement learning algorithms to predict and reconstruct the optimal processing route and time allocation.

[0034] The current batch's flexible bill of materials attribute range, real-time capacity data of each participating entity, and logistics delay risk factors are normalized and concatenated to form a reinforcement learning state vector, denoted as the state at decision time t. The action space is defined as a set of adjustments to the process route, including adding and deleting process nodes, adjusting the time allocation of each process node, and reallocating processing entities. These actions are a mixture of discrete and continuous actions. A reward function is constructed based on indicators such as the total project duration, resource utilization rate, and quality pass rate after the process route is executed. Positive reward values ​​are assigned to actions that shorten the project duration, improve resource utilization rate, and increase quality pass rate, while negative penalty values ​​are assigned to actions that cause resource conflicts or quality decline.

[0035] The proximal policy optimization algorithm is used to iteratively update the policy network parameters. Its pruning objective function is expressed as:

[0036] ;

[0037] The probability ratio of the old and new strategies on the same action is expressed as:

[0038] ;

[0039] In the above formula, This represents the pruning objective function of the near-end policy optimization algorithm. This represents the expected estimate for time step t. This indicates the difference between the old and new strategies in action. The probability ratio on This indicates that the updated policy is in the state. Select action The probability, This indicates the state of the old strategy before the update. Select action The probability, This represents the advantage function estimate at time step t, used to measure the performance of the action relative to the average level. This represents a clipping function that restricts the probability ratio to a preset range. This represents the pruning magnitude hyperparameter, used to limit the magnitude of a single policy update to ensure training stability.

[0040] The converged policy network is applied to the current batch status to output process route adjustment actions. For example, when a batch is found to have a high moisture content range, a drying milestone node is inserted into the process route and the corresponding working hours are reallocated. The reconstructed process route is parsed into a standard format and pushed to the project planning module of the product data management system, triggering the automatic update of project task nodes and schedule plans. The actual execution time, capacity utilization and other feedback data collected after actual execution are re-included in the experience buffer. The policy network is incrementally retrained periodically so that the process route reconstruction capability is continuously optimized as production data accumulates.

[0041] In another implementation, the dynamic reconfiguration of the process route can also employ other policy gradient-based reinforcement learning algorithms, or combine graph neural networks to model the topological relationships of multi-entity production capacity, replacing the aforementioned near-end policy optimization algorithm. By adjusting the state space representation and reward function design, dynamic reconfiguration of the process route and time allocation can also be achieved.

[0042] Example 4: Automatic alignment of five-stream convergence milestones using a time-series knowledge graph

[0043] This mechanism solves the problems of data silos among multiple entities in the supply chain and the distortion of progress caused by manual reporting. Its core is to build a time-series knowledge graph to perform entity alignment and credibility scoring on heterogeneous data, thereby replacing manual reporting and achieving automatic milestone updates.

[0044] The big data platform extracts raw event data related to the current project from the commercial flow system, logistics system, information flow system, business flow system, and capital flow system through interface connections. The commercial flow system corresponds to order and transaction records, the logistics system corresponds to transportation and warehousing records, the information flow system corresponds to records filled in by various entities or reported by sensors, the business flow system corresponds to processing and quality inspection records, and the capital flow system corresponds to settlement and payment records.

[0045] Entity recognition is performed on the extracted heterogeneous event data. Cross-system entities describing the same batch and the same project node are aligned. The similarity between entities is calculated as follows:

[0046] ;

[0047] in, Representing entities With entity Similarity score between them , Representing entities respectively , The feature vector is obtained based on its attribute text or structured field encoding. Represents the dot product of two vectors. , The two vectors represent their magnitudes respectively. When the similarity score exceeds a preset threshold, the two entities are determined to point to the same business object. They are then merged into the same node in the knowledge graph and their timestamps in their respective data sources are labeled to form a temporal relationship.

[0048] For duplicate or conflicting records of the same entity appearing in multiple data sources, a weighted cross-validation method is used to calculate the credibility score of the entity's data:

[0049] ;

[0050] in, This represents the credibility score of the i-th data record to be evaluated. This indicates the total number of other data source records that are related to this record. This represents the credibility weight of the j-th associated data source, which is preset based on the historical accuracy of the data source. Indicates the data record to be evaluated Related data records Consistency and similarity between them.

[0051] When the credibility score is lower than the preset threshold, the corresponding data record is marked as low-credibility data, triggering a manual review prompt. When the credibility score reaches the preset threshold, the corresponding data record is determined to be valid data and enters the subsequent mapping process. The filtered valid data is automatically mapped to the corresponding milestone node in the project plan of the product data management system according to its time sequence and entity alignment results. The progress status, completion time, and associated cost and quality data of the node are updated, replacing the traditional manual entry method and realizing the automatic and accurate integration of project management progress and underlying business chain flow.

[0052] The flexible bill of materials dynamic instantiation mechanism, cross-entity process route dynamic reconstruction mechanism, and five-stream integration milestone automatic alignment mechanism involved in the above embodiments, through the coordinated operation of a unified big data platform and the connection with existing commercial product data management systems via standardized interfaces, together constitute the complete technical solution for which this invention seeks protection. This solution can solve the technical problems of unstructured bill of materials expression, difficulties in cross-entity collaborative production scheduling, and distortion of project progress tracking in agricultural and sideline product project management.

Claims

1. A method for managing agricultural and sideline product projects based on big data technology, characterized in that: include: Collect multidimensional spatiotemporal natural attribute raw data of agricultural and sideline product raw materials when they arrive at the time of delivery, and preprocess the multidimensional spatiotemporal natural attribute raw data to obtain a standardized dataset; Feature vectors are constructed in batches based on the standardized dataset. A Gaussian mixture model is used to cluster the feature vector set composed of multiple batches of feature vectors to obtain multiple clusters. For each cluster, an attribute fluctuation range is generated based on the mean and standard deviation of the samples in each attribute dimension. The attribute fluctuation range is used as the fuzzy master data level corresponding to the cluster. The feature vector of the batch to be processed is assigned to the cluster with the highest posterior probability. Based on the fuzzy master data level corresponding to the cluster, a flexible bill of materials entry for the batch to be processed is generated. A version number is assigned to the flexible bill of materials entry. The version number is bound to the batch traceability code, the collection timestamp, and the original data of the multidimensional spatiotemporal natural attributes and then stored in the database. The flexible bill of materials entry is then pushed to the material master data module of the product data management system. Using the attribute fluctuation range of the flexible bill of materials items, the real-time capacity data of each participant in the supply chain, and the logistics delay risk factor as state inputs, a reinforcement learning state vector is constructed. The policy network outputs process route adjustment actions based on the state vector. The process route adjustment actions include at least one of adding and deleting process nodes, adjusting the time allocation of process nodes, and reallocating processing entities. A reward function is constructed based on the total project duration data, resource utilization rate data, and quality pass rate data after the process route is executed. The network parameters of the strategy network are iteratively updated based on the reward function to obtain the updated strategy network. The updated strategy network outputs the process route adjustment action corresponding to the current batch's state vector and pushes it to the project planning module of the product data management system. Raw event data related to the current project is extracted from the commercial flow system, logistics system, information flow system, business flow system, and capital flow system respectively. Entity identification is performed on the raw event data, cross-system entities pointing to the same batch and the same project node are aligned, and the timestamps of each cross-system entity in the corresponding data source are marked to form a time-series knowledge graph. For data records of the same entity appearing in multiple data sources, a credibility score for the data record is calculated based on the consistency and similarity between the data record and related data records, as well as the credibility weight of the related data sources. Data records whose credibility scores meet the preset threshold conditions are mapped to the corresponding milestone nodes in the project plan of the product data management system according to their temporal relationship and entity alignment results. The progress status, completion time, and associated cost and quality data of the milestone node are updated, and the updated milestone node data is fed back to the state vector construction stage of the reinforcement learning for the output of the next batch of process route adjustment actions.

2. The method according to claim 1, characterized in that, Before clustering the feature vector set using a Gaussian mixture model, the method further includes: evaluating the Gaussian mixture model under each candidate cluster number using the Bayesian information criterion to obtain the evaluation index corresponding to each candidate cluster number. The evaluation index is determined based on the log-likelihood function value of the model under the corresponding candidate cluster number, the number of free parameters of the model, and the total number of batch samples participating in the clustering. The candidate cluster number that meets the preset optimization conditions is selected as the cluster number.

3. The method according to claim 1, characterized in that, The attribute fluctuation range is obtained by adding and subtracting the product of a preset interval width coefficient and the standard deviation of the mean of the samples in the corresponding attribute dimension within the cluster. The preset interval width coefficient is a configurable value.

4. The method according to claim 1, characterized in that, The iterative update of the network parameters of the policy network based on the reward function includes: constructing a pruning objective function based on the probability ratio of the old and new policies on the same action and the advantage function estimate, limiting the probability ratio within a preset range, and performing gradient updates of the network parameters of the policy network based on the pruning objective function.

5. The method according to claim 1, characterized in that, Also includes: The project schedule data and capacity utilization data collected after the process route is executed are included in the experience buffer, and the strategy network is incrementally retrained based on the data in the experience buffer.

6. The method according to claim 1, characterized in that, The entity recognition process for the original event data, which aligns cross-system entities pointing to the same batch and the same project node, includes: obtaining the feature vector of the corresponding entity based on the attribute text or structured field encoding of the entity; calculating the similarity score between the two entities based on the inner product of the feature vectors corresponding to each entity and the magnitude of the feature vectors; and determining that the two entities point to the same business object when the similarity score exceeds a preset threshold, and merging the two entities into the same node in the time-series knowledge graph.

7. The method according to claim 1, characterized in that, When the credibility score is lower than a preset threshold, the corresponding data record is marked as low-credibility data, and a manual review prompt is triggered.

8. A project management system for agricultural and sideline products based on big data technology, characterized in that, include: The data acquisition module is used to collect multi-dimensional spatiotemporal natural attribute raw data of agricultural and sideline product raw materials when they arrive, as well as production capacity data, logistics data, and business flow data of various participants in the supply chain; The flexible bill of materials generation module is used to preprocess the original data of the multidimensional spatiotemporal natural attributes and construct feature vectors. It uses a Gaussian mixture model to cluster the feature vector set to obtain multiple clusters. It generates attribute fluctuation ranges as fuzzy master data levels based on the mean and standard deviation of samples within the clusters, and generates flexible bill of materials entries based on the fuzzy master data levels. The process route dynamic reconstruction module is used to take the attribute fluctuation range of the flexible bill of materials item, the real-time capacity data of the participating entity and the logistics delay risk factor as status input, output process route adjustment actions through the strategy network, and iteratively update the network parameters of the strategy network based on the feedback data after the process route is executed. The milestone alignment module is used to extract raw event data from the commercial flow system, logistics system, information flow system, business flow system, and capital flow system respectively, construct a time-series knowledge graph for entity alignment, calculate the credibility score of data records, and map data records that meet the preset threshold conditions to the corresponding milestone nodes in the project plan. The standard interface module is used to parse the flexible bill of materials entries, the process routes corresponding to the process route adjustment actions, and the update results of the milestone nodes into a standardized data format and push them to the product data management system.

9. The system according to claim 8, characterized in that, The data acquisition module includes at least one of a multispectral detection device, a smart weighbridge, an interface to a local weather station, and an IoT sensor, used to collect the raw data of the multidimensional spatiotemporal natural attributes.

Images