Sorghum whole industry chain data management method and system

Abstract

The invention provides a sorghum whole industry chain data management method and system, and relates to the technical field of Internet of Things edge computing, and the method comprises the steps: analyzing the multi-mode fuzzy features of the target data of a sorghum whole industry chain collected through the Internet of Things; based on the multi-modal fuzzy features, matching an optimal edge computing node for the target data; and transmitting the target data to the optimal edge calculation node for edge calculation to obtain an edge calculation result, and outputting the edge calculation result. According to the sorghum whole industry chain data management method and system, the edge computing technology is effectively integrated into the sorghum whole industry chain data management system, the data does not need to be stored in the same data center, the delay problem possibly existing in data transmission is avoided, the data does not need to be analyzed and applied through manual operation, and the data management efficiency is improved. And the data management efficiency is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things edge computing technology, and in particular to a data management method and system for the entire sorghum industry chain. Background Technology

[0002] The sorghum industry chain involves multiple stages, from planting and processing to transportation and sales. During this process, data collection via the Internet of Things (IoT) generates a massive amount of data, which is crucial for monitoring sorghum growth, supporting planting decisions, controlling processing quality, and managing product sales. However, this data is typically stored centrally in a single data center, potentially leading to data transmission delays. Furthermore, the analysis and application of this data currently rely on manual intervention, further reducing management efficiency.

[0003] As edge computing technology matures, data processing tasks can be shifted from traditional data centers to edge nodes closer to the data source. This approach enables real-time data processing and analysis near the source, effectively reducing transmission latency and improving real-time performance and computational efficiency. This technology has significant application potential in data management across the entire sorghum industry chain.

[0004] Therefore, how to effectively integrate edge computing technology into the data management system of the entire sorghum industry chain to overcome the current problems of latency and management efficiency is an urgent issue to be addressed. Summary of the Invention

[0005] One of the objectives of this invention is to provide a data management method for the entire sorghum industry chain to solve the problems in the background art.

[0006] This invention provides a data management method for the entire sorghum industry chain, comprising: Analyze the multimodal fuzzy features of target data from the entire sorghum industry chain collected through the Internet of Things; Based on multimodal fuzzy features, the optimal edge computing nodes are matched for the target data; The target data is transmitted to the optimal edge computing node for edge computing, and the edge computing results are obtained and output.

[0007] Optionally, the multimodal fuzzy features include at least: data type, data source type, and spatiotemporal information.

[0008] Optionally, the step of matching the optimal edge computing node for the target data based on multimodal fuzzy features includes: Predict the edge computing needs of target data reflected by multimodal fuzzy features; wherein the types of edge computing needs include at least: sorghum growth monitoring, planting decision support, processing quality control, and product sales management; The candidate edge computing node whose currently available edge computing capacity best meets the edge computing requirements is selected as the optimal edge computing node and matched with the target data.

[0009] Optionally, when transmitting the target data to the optimal edge computing node for edge computing, if the optimal edge computing node reports that additional supplementary data is needed to support the continued edge computing of the target data, and the supplementary data exceeds the range of IoT acquisition capabilities, the supplementary data is optimized so that the acquisition difficulty of the optimized supplementary data and the support strength of the optimal edge computing node for the continued edge computing of the target data achieve the optimal balance. Based on the data acquisition capabilities, the scheduling network is used to supplement the specific content of the optimized supplementary data to the maximum extent. The additional data collected will be transmitted as feedback to the optimal edge computing node.

[0010] Optionally, the optimization of the supplementary data aims to achieve an optimal balance between the difficulty of collecting the optimized supplementary data and the support strength of the optimal edge computing node for further edge computing on the target data, including: Plan an N-level optimization scheme to supplement the data; Take i from 1 to N in sequence, and determine the value m of i when the difficulty of collecting the supplementary data after optimization by the i-th level optimization scheme and the support strength of the edge computing node for the target data to continue edge computing are closest to the balance target. The supplementary data optimized by the m-level optimization scheme is selected as the final optimized supplementary data. The balance objectives include: The difficulty of data collection is set at the first equilibrium threshold. The level of support is the second equilibrium threshold.

[0011] Optionally, the N-level optimization scheme for the planning supplementary data includes: Based on the data type distribution of the supplementary data, suitable optimization scheme planning rules for optimizing the supplementary data are determined from the optimization scheme planning rule base; Based on the optimization scheme planning rules, plan an N-level optimization scheme for supplementary data.

[0012] Optionally, the steps for obtaining the first balance threshold and the second balance threshold include: The analysis supplements the upper limit of the collection difficulty represented by the collection capability; The lower limit of the support level represented by the degree of supplementary data required to analyze the optimal edge computing node; Based on the relationship between the upper limit of difficulty and the lower limit of intensity, a balance rule is determined from the balance rule base to achieve the optimal balance between the collection difficulty and the collection intensity. The balance rule includes: the upper limit adjustment coefficient and the lower limit adjustment coefficient. Based on the upper limit reduction coefficient, the upper limit of difficulty is reduced to obtain the first balance threshold; Based on the lower limit adjustment coefficient, the lower limit of intensity is adjusted upward to obtain the second equilibrium threshold.

[0013] Optionally, the specific content of the supplementary data after collection optimization based on the data acquisition capability mining scheduling network includes: In the data acquisition capability mining and scheduling network, the path that should be generated for the optimized supplementary data is represented; Based on the network environment information of the path that should be generated, the target local path that should be generated is determined to start generating key feature states and develops on the optimized supplementary data on the path that should be generated. Acquire the path segments of multiple data acquisition nodes that will pass through the target local path in the future, and the changes in their acquisition capabilities as they pass through the target. Based on the path segments of each data acquisition node and the changes in data acquisition capabilities during the path, we explore the relay data acquisition capabilities of each data acquisition node to continuously attempt to collect specific content from the optimized supplementary data. Based on relay data acquisition capabilities, each acquisition node is scheduled to continuously attempt to acquire the specific content of the optimized supplementary data.

[0014] Optionally, the relay acquisition capability, which involves mining the relay acquisition capability of each acquisition capability node to continuously attempt to acquire specific content from the optimized supplementary data based on the path segments of each acquisition capability node and the changes in acquisition capability during the path, includes: Each data acquisition node is divided into multiple node sets; in which, the path segments of multiple data acquisition nodes in the same node set overlap in pairs. Iterate through each node set in turn; Each time a node is traversed, the relay collection capacity requirement is reflected by the sum of the path segments of each collection capacity node in the traversed node set. To address the need for relay data collection capabilities, we attempted to identify potential relay data collection capabilities by exploring the changes in data collection capabilities of each data collection capability node during the traversal of the node set. After traversing each node set in sequence, the candidate relay collection capabilities that were successfully mined during each traversal are reduced and optimized to obtain relay collection capabilities.

[0015] This invention provides a sorghum whole-industry chain data management system, comprising: The feature analysis module is used to analyze the multimodal fuzzy features of target data from the entire sorghum industry chain collected through the Internet of Things; The node matching module is used to match the optimal edge computing node for the target data based on multimodal fuzzy features; The edge computing module is used to transmit target data to the optimal edge computing node for edge computing, obtain edge computing results, and output them.

[0016] The present invention has achieved the following beneficial effects: When the Internet of Things collects target data from the entire sorghum industry chain, its multimodal fuzzy characteristics are analyzed, and the optimal edge computing node is immediately matched for edge computing to realize sorghum growth monitoring, planting decision support, processing quality control, and product sales management. This effectively integrates edge computing technology into the data management system of the entire sorghum industry chain, eliminating the need to store data in the same data center, avoiding potential data transmission delays, and eliminating the need for manual data analysis and application, thus greatly improving data management efficiency.

[0017] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of a sorghum whole-industry chain data management method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a sorghum whole-industry chain data management system according to an embodiment of the present invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] Example 1: This invention provides a data management method for the entire sorghum industry chain, such as... Figure 1 As shown, it includes: S100. Analyze the multimodal fuzzy features of target data from the entire sorghum industry chain collected through the Internet of Things.

[0022] The target data includes at least: sorghum growth environment data, growth status data, related agricultural equipment status data, processing status data, logistics and transportation data, market sales data, etc., which can be collected by relevant IoT sensors pre-deployed at different nodes of the industrial chain; the multimodal fuzzy feature is the feature that the target data has fuzziness in multiple modes.

[0023] S200: Based on multimodal fuzzy features, it matches the optimal edge computing node for the target data.

[0024] Among them, the optimal edge computing node is the most suitable edge computing node for performing edge computing on target data. It is preset based on edge computing technology and pre-configured with relevant edge computing capabilities.

[0025] S300: Transmit the target data to the optimal edge computing node for edge computing, obtain the edge computing results, and output them.

[0026] Among them, the optimal edge computing node performs edge computing on the target data to realize sorghum growth monitoring, planting decision support, processing quality control and product sales management, and outputs the final edge computing results.

[0027] The above-mentioned technical solution analyzes the multimodal fuzzy characteristics of the target data collected by the Internet of Things across the entire sorghum industry chain, and immediately matches the optimal edge computing node for edge computing to realize sorghum growth monitoring, planting decision support, processing quality control, and product sales management. This effectively integrates edge computing technology into the data management system of the entire sorghum industry chain, eliminating the need to store data in the same data center, avoiding potential data transmission delays, and eliminating the need for manual data analysis and application, thus greatly improving data management efficiency.

[0028] Example 2: In this embodiment of the invention, the multimodal fuzzy features include at least: data type, data source type, and spatiotemporal information.

[0029] Spatiotemporal information refers to the temporal characteristics (such as data acquisition time) and spatial characteristics (such as data acquisition location) of the target data. Specifically, the analysis of multimodal fuzzy features is achieved by constructing a fuzzy membership matrix. The system pre-sets a set of standard feature vectors for each link in the sorghum industry chain. ,in Representative cluster centers are generated for standard states such as "seedling stage - water shortage" and "grain-filling stage - disease". These cluster centers are obtained by pre-training and extracting data using K-Means or DBSCAN clustering algorithms based on a historical big data database of the entire sorghum industry chain. This is for target data collected via the Internet of Things. The input vector (containing data type encoding, source credibility score, and normalized spatiotemporal coordinates) is constructed, and the target data is calculated using the fuzzy C-means (FCM) algorithm. For each standard feature membership degree The membership vector is the multimodal fuzzy feature, which mathematically describes the degree of fuzzy transition between "normal" and "abnormal" states of data. Specifically, the source credibility score is a quantitative indicator pre-set based on the hardware accuracy and stability of the data acquisition equipment, preferably ranging from 0 to 1. In this embodiment: for fixed high-precision industrial sensors (such as weather stations), the score is set to 0.9~1.0; for mobile automated acquisition equipment (such as drones), the score is set to 0.7~0.9; for manually entered data, the score is set to 0.5~0.7. This hierarchical setting ensures that high-precision data sources have higher weights when calculating multimodal features.

[0030] Example 3: In this embodiment of the invention, step S200, matching the optimal edge computing node for the target data based on multimodal fuzzy features, includes: S2001. Predict the edge computing needs of target data reflected by multimodal fuzzy features; wherein the types of edge computing needs include at least: sorghum growth monitoring, planting decision support, processing quality control, and product sales management.

[0031] Among them, multimodal fuzzy features reflect the edge computing needs of the target data. For example, if the multimodal fuzzy features are data on the growth environment of sorghum and the data comes from planting area A, then the edge computing need reflected is to analyze whether the growth environment of planting area A is suitable.

[0032] S2002. Match the candidate edge computing node whose currently available edge computing capabilities best meet the edge computing requirements with the target data as the optimal edge computing node.

[0033] The most suitable specific calculation logic is as follows: Let the predicted edge computing demand feature vector be... (Representing the required CPU computing power, memory, and specific sorghum growth model version, respectively), candidate edge computing nodes The idle capacity feature vector is .

[0034] Calculate the matching score : in, For nodes Network latency between the data source and the data source Weighting coefficients; subscript This indicates normalization processing, specifically... ,in This represents the maximum CPU computing power among all candidate edge computing nodes (the same applies to memory). This ensures that capability parameters of different dimensions are mapped to the [0,1] interval for uniform weighted calculation. Specifically, to prevent the denominator from being 0 and to ensure numerical stability, the normalized calculation formula is modified as follows: ,in The minimum constant (take) Furthermore, for the formula... The unit is uniformly set to milliseconds (ms), and normalization processing is required before substituting into the formula. The processing method is as follows: If normalization is not performed, then weighting is required. Perform dimensional balancing. (For model version) If the candidate node does not support this version, then the matching degree of that node will be directly reduced. Select The node with the highest value that meets the basic threshold is selected as the optimal edge computing node. It should be noted that in the formula... For the weighting coefficients, satisfying Those skilled in the art can adjust its value according to actual business needs: if the task is computationally intensive (such as disease image recognition), focusing on CPU and memory, it is preferable to set it to... If the task is latency-sensitive (such as real-time logistics tracking), and low latency is the priority, it is best to set it to... In the absence of special preferences, all coefficients are assumed to be 1 / 3 by default.

[0035] Among them, candidate edge computing nodes are available edge computing nodes configured with edge computing capabilities. Idle edge computing capabilities refer to the portion of capabilities that are not currently used. The candidate edge computing node whose currently idle edge computing capabilities best meet the edge computing requirements (such as being able to analyze whether the growth environment of planting area A is suitable and being close to the data source of planting area A) is selected as the optimal edge computing node, which is most suitable for edge computing analysis of the target data. When pre-configuring edge computing capabilities, a data analysis application model for the entire sorghum industry chain can be pre-trained (such as using a large amount of historical sorghum growth anomaly data as training samples for machine learning training to obtain a sorghum growth monitoring model), and integrated onto nodes deployed based on edge computing technology.

[0036] The above technical solution predicts the edge computing needs of the target data reflected by the multimodal fuzzy features, and selects the candidate edge computing node with the most suitable edge computing capabilities as the optimal edge computing node to match with the target data, thereby improving the accuracy and efficiency of matching the target data with the optimal edge computing node.

[0037] Example 4: The entire sorghum industry chain involves numerous growing, processing, transportation, and sales sites, making it difficult to comprehensively cover all aspects of IoT data collection. As the edge computing capabilities of edge computing nodes iterate or undergo autonomous reinforcement learning, when performing deeper edge computing (such as in-depth data analysis applications for sorghum growth monitoring, planting decision support, processing quality control, and product sales management), there may be a lack of supplementary data beyond the scope of IoT data collection. If optimal supplementary data collection and processing cannot be performed in this case, the potential of edge computing nodes will not be fully realized, affecting the development of data management across the entire sorghum industry chain.

[0038] Therefore, in this embodiment of the invention, the sorghum whole-industry chain data management method further includes: S400 When transmitting the target data to the optimal edge computing node for edge computing, if the optimal edge computing node reports that additional supplementary data is needed to support the continued edge computing of the target data, and the supplementary data exceeds the range of IoT acquisition capabilities, the supplementary data is optimized to achieve an optimal balance between the difficulty of acquiring the optimized supplementary data and the support strength of the optimal edge computing node for the continued edge computing of the target data.

[0039] The IoT data collection capability refers to the range of data that can be collected across the entire sorghum industry chain through the IoT. Collection difficulty refers to the difficulty of collecting the optimized supplementary data, while support strength refers to the level of support the optimized supplementary data provides to the optimal edge computing nodes for further, deeper edge computing on the target data. To ensure the smooth collection of the optimized supplementary data, the collection difficulty needs to be minimized. However, to ensure the optimal edge computing nodes' ability to continue edge computing on the target data, the support strength needs to be maximized. Therefore, when optimizing the supplementary data, an optimal balance needs to be struck between collection difficulty and support strength.

[0040] S500, based on the acquisition capability, mines and schedules the network to maximize the supplementary data content after acquisition optimization; The data acquisition capability mining and scheduling network is a pre-defined network used to mine and schedule the specific content of the supplementary data after optimization. Based on this network, the specific content of the optimized supplementary data is supplemented to the maximum extent.

[0041] S600 transmits the supplementary collected data as feedback to the optimal edge computing node.

[0042] The specific content collected is transmitted as feedback to the optimal edge computing node, which can then perform deeper edge computing on the target data based on this specific content.

[0043] The above technical solution addresses the issue that the optimal edge computing node requires additional supplementary data to support further edge computing of the target data. When this supplementary data exceeds the IoT's data collection capabilities, the supplementary data is optimized. This achieves an optimal balance between the difficulty of collecting the optimized supplementary data and the support provided by the optimal edge computing node for further edge computing of the target data. Based on the data collection capability mining and scheduling network, the specific content of the optimized supplementary data is collected to the maximum extent and transmitted as feedback to the optimal edge computing node. This ensures optimal supplementary data collection processing, avoids underutilization of the edge computing node's potential, improves the data management level of the entire sorghum industry chain, and further effectively integrates edge computing technology into the data management system of the entire sorghum industry chain.

[0044] Example 5: In this embodiment of the invention, in step S400, the supplementary data is optimized to achieve an optimal balance between the difficulty of collecting the optimized supplementary data and the support strength of the optimal edge computing node for continuing edge computing on the target data, including: S4001, N-level optimization scheme for planning supplementary data.

[0045] First, N levels of optimization schemes are planned for the supplementary data. Each level of optimization scheme implements different optimization measures for the supplementary data. The higher the level, the greater the changes to the supplementary data caused by the optimization measures implemented by the corresponding optimization scheme.

[0046] S4002. Take i from 1 to N in sequence, and determine the value m of i when the difficulty of collecting the supplementary data after optimization by the i-th level optimization scheme and the support strength of the edge computing node for the target data to continue edge computing are closest to the balance target.

[0047] The quantification steps for the acquisition difficulty of the supplementary data after optimization by the i-th level optimization scheme and the support strength of the optimal edge computing node for continued edge computing of the target data include: weighting the complexity of the optimized data after optimization by the i-th level optimization scheme and the total number of times related data of the optimized data has been successfully acquired in the past (the weights of the weighting can be pre-set according to the degree of influence of complexity and total number of acquisitions on the acquisition difficulty), to obtain the acquisition difficulty of the optimized data; the specific quantification of acquisition difficulty is defined by the formula: ,in For the first Level of data volume Weighting coefficients are applied to the preset data complexity. This is an environmental impact factor determined based on the number of successful supplementary data collections in history (the higher the number of successful historical data collections, the smaller the value of this factor). As a specific implementation method, A weighting factor for data complexity, used to eliminate the influence of units, and its preferred value is... to (corresponding data volume) (When measured in KB). The specific calculation logic is as follows ,in This represents the number of successful data collections in history. It is a natural constant; therefore, it can be seen that with historical data collection experience... The increase in environmental impact factors It will gradually decrease, thereby reducing The value is used to characterize the decrease in data collection difficulty. Furthermore, regarding the amount of data in the formula... Its unit of measurement is set to kilobytes (KB). Coefficient As a unit conversion coefficient for transforming data volume into difficulty scores, its value range is set to be... to For the natural constant The value is approximately 2.718, used to prevent errors due to historical data collection counts. When the logarithmic value is 0, the logarithmic function is meaningless.

[0048] Next, the level of support is quantified: the supplementary data is processed through the [missing information - likely a process or procedure]. The degree to which the optimized data meets the needs of the optimal edge computing node for supplementary data (such as the clarity requirement of sorghum lesion texture) and the quality of the optimized data are weighted and calculated to obtain the support strength of the optimized data.

[0049] Specifically, the N-level optimization scheme refers to hierarchical precision reduction or mode transformation processing of the data. For example: Level 1 (i=1) involves acquiring full-band hyperspectral images (largest data volume, strongest support); Level 2 (i=2) involves acquiring RGB visible light images; Level N (i=N) involves acquiring only manually described text data (smallest data volume, weakest support).

[0050] Support The calculation formula is: ,in For information entropy, For model confidence, This is a support strength adjustment coefficient used to balance information entropy and model confidence. In this embodiment, The optimal value is a constant between 1.0 and 5.0. When the system strategy is conservative (i.e., requires extremely high data support to trigger edge computing), the value can be lowered. The value is adjusted upwards; conversely, it is adjusted upwards. The Shannon entropy formula, based on data distribution, can be used to calculate the information richness of data. The specific formula is as follows: ,in To supplement the feature values ​​in the data (such as the gray level of the image or the sensor value). This represents the probability of the feature value appearing in the data distribution. This metric measures the amount of information carried by the data; a higher entropy value indicates that the data contains more non-redundant information and has greater potential to support edge computing. This represents the prediction confidence probability (0~1) of the pre-trained model for this level of data input.

[0051] The final equilibrium objective function is: .

[0052] Next, a weighted calculation is performed on the degree to which the optimized data, after being optimized by the i-th level optimization scheme, meets the requirements of the optimal edge computing node for supplementary data (such as the type and accuracy of the supplementary data) and the quality of the optimized data (determined by a comprehensive evaluation of reliability, completeness, etc., which can be pre-set using a quantification system based on these properties). The weights of this weighted calculation can be pre-set according to the degree to which the requirements are met and the quality of the data influences the support level. This yields the acquisition difficulty of the optimized data. The balance target refers to achieving a balance between acquisition difficulty and support level. The value of i, m, is determined when the acquisition difficulty and support level of the supplementary data, after being optimized by the i-th level optimization scheme, are closest to the balance target.

[0053] S4003. Select the supplementary data optimized by the m-level optimization scheme as the final optimized supplementary data.

[0054] Among them, the supplementary data optimized by the m-level optimization scheme is selected as the final optimized supplementary data. Its collection difficulty and support strength are closest to the balance target, so that the collection difficulty of the optimized supplementary data and the support strength of the ability of the optimal edge computing node to continue edge computing on the target data reach the optimal balance.

[0055] The balance objectives include: The difficulty of data collection is set at the first equilibrium threshold. The level of support is the second equilibrium threshold.

[0056] In the balancing objective, a first balancing threshold and a second balancing threshold are set for when the difficulty of data collection and the level of support are balanced.

[0057] The above-mentioned pre-set balance target and planned N-level optimization schemes for supplementary data, taking i from 1 to N sequentially, determined the value m of i when the collection difficulty and support strength of the supplementary data after optimization by the i-th level optimization scheme are closest to the balance target. The supplementary data optimized by the m-level optimization scheme is selected as the final optimized supplementary data. This greatly improves the accuracy, comprehensiveness and efficiency of the optimization of supplementary data, and enhances the ability to achieve the optimal balance between the collection difficulty of the optimized supplementary data and the support strength of the optimal edge computing node for the target data to continue edge computing.

[0058] Example 6: In this embodiment of the invention, the N-level optimization scheme for the planning supplementary data includes: Based on the data type distribution of the supplementary data, suitable optimization scheme planning rules for optimizing the supplementary data are determined from the optimization scheme planning rule base; Based on the optimization scheme planning rules, plan an N-level optimization scheme for supplementary data.

[0059] This involves a pre-set optimization scheme planning rule base, containing optimization scheme planning rules corresponding to different data type distributions of supplementary data. The rules are directly retrieved from the base base, and then an N-level optimization scheme for the supplementary data is planned based on these rules. Data type distribution refers to the multiple data types of the supplementary data. Specifically, for example, the importance of each data type can be analyzed based on the data type distribution. When setting optimization scheme planning rules, the higher the level, the more data types with lower importance are removed from the corresponding optimization scheme. Furthermore, measures such as modifying the accuracy and error requirements of the supplementary data can be set as optimization methods for the optimization scheme.

[0060] Example 7: In this embodiment of the invention, the steps for obtaining the first balance threshold and the second balance threshold include: The analysis supplements the upper limit of the collection difficulty represented by the collection capability.

[0061] Among them, the supplementary data collection capability refers to the scope of specific content that the system may be able to collect from supplementary data, and the upper limit of the collection difficulty it represents refers to the maximum collection difficulty it can accept.

[0062] The degree to which the optimal edge computing node needs supplementary data represents the lower limit of the support strength.

[0063] The degree to which the optimal edge computing node needs supplementary data refers to the degree of its need for supplementary data, which can be determined from the feedback of the optimal edge computing node. The lower limit of the support strength it represents refers to the minimum degree of support the optimal edge computing node needs for supplementary data. Based on the relationship between the upper limit of difficulty and the lower limit of support strength, a balancing rule is determined from the balancing rule base to achieve an optimal balance between the collection difficulty and the data acquisition difficulty. This balancing rule includes: an upper limit adjustment coefficient and a lower limit adjustment coefficient.

[0064] The balancing rule library pre-sets balancing rules corresponding to the optimal balance between the upper and lower limits of difficulty, ensuring a balance between the two. These rules are determined directly by querying the library. Each balancing rule includes an upper limit adjustment coefficient and a lower limit adjustment coefficient, where the upper limit adjustment coefficient is less than or equal to 1, and the lower limit adjustment coefficient is ≥1. Based on the upper limit adjustment coefficient, the upper limit of difficulty is lowered to obtain the first balancing threshold.

[0065] Specifically, when the upper limit of difficulty is lowered based on the upper limit reduction coefficient, the upper limit of difficulty is multiplied by the upper limit reduction coefficient to obtain the first balance threshold.

[0066] Based on the lower limit adjustment coefficient, the lower limit of intensity is adjusted upward to obtain the second equilibrium threshold.

[0067] Specifically, when adjusting the lower limit of intensity upward based on the lower limit adjustment coefficient, the lower limit of intensity is multiplied by the lower limit adjustment coefficient to obtain the second equilibrium threshold.

[0068] The mathematical expression corresponding to the above logic is as follows: Let the upper limit of the collection difficulty be... The lower limit of the support level is .

[0069] First equilibrium threshold The calculation formula is: ,in This is the upper limit adjustment factor, and its value range is [value range missing]. ; Second equilibrium threshold The calculation formula is: ,in This is the adjustment factor for the lower limit, and its value range is... .

[0070] By introducing coefficients and The system can dynamically shrink the solution space while ensuring minimum requirements, thereby accelerating the process to the optimal equilibrium point. The search converged.

[0071] The above technical solution dynamically analyzes and supplements the upper limit of the acquisition difficulty represented by the acquisition capability and the lower limit of the support strength represented by the degree of supplementary data required by the optimal edge computing node. Based on the relationship between the upper limit of difficulty and the lower limit of strength, the balance rule that achieves the optimal balance between acquisition difficulty and acquisition capability is determined from the balance rule library. The upper limit of difficulty is lowered and the lower limit of strength is raised using the balance rule, and finally the first balance threshold and the second balance threshold are obtained. This realizes the real-time dynamic setting of the optimal balance target, which greatly improves the accuracy of its use in determining the optimized supplementary data.

[0072] Example 8: In this embodiment of the invention, the specific content of S500, which involves using a data acquisition capability-based scheduling network to maximize the supplementary data after acquisition optimization, includes: S5001, In the data acquisition capability mining and scheduling network, the optimized supplementary data should be represented by its intended path. This data acquisition capability mining and scheduling network is constructed as a dynamic spatio-temporal graph. ,in This refers to fixed sensors and mobile data acquisition terminals (such as drones and inspection vehicles) across the entire industry chain. Represents a physically reachable path. This represents the set of time slices for the movement and acquisition tasks of the data acquisition node.

[0073] The path should be a predicted trajectory mapping generated based on the production plan and logistics scheduling table of the sorghum lifecycle management (PLM) system.

[0074] The key feature state is defined as: when data features (such as leaf yellowing rate) exceed a preset abnormal threshold, triggering the edge computing model to intervene deeply.

[0075] The data acquisition capability mining and scheduling network includes 3D maps of the entire sorghum industry chain, involving numerous growth sites, processing sites, transportation sites, and sales sites. It also includes the future data acquisition paths of different data acquisition nodes within these sites (such as data acquisition vehicles equipped with RFID readers, cameras, and mobile personnel). Correspondingly, this includes the future movement paths of data acquisition vehicles and personnel executing their work plans. The optimized supplementary data generation path refers to the path that the optimized supplementary data should sequentially pass through different stages from its initial generation to its final formation. For example, if the optimized supplementary data is the state of sorghum planted in an experimental area undergoing a pest and disease resistance experiment, with precise comparisons of plant height, leaf discoloration, and pest and disease traces, then the expected generation path would be: starting with sorghum planting in the experimental area, entering the experimental area for initial monitoring of the sorghum, collecting basic growth data (such as plant height and leaf health status), then conducting a detailed visual inspection of the sorghum in the experimental area, recording the state of each sorghum plant, especially detailed observation of leaf discoloration and pest and disease traces, and finally comparing the data.

[0076] S5002. Based on the network environment information of the path that should be generated, determine the target local path on the path that should be generated, where the optimized supplementary data begins to generate key feature states and develops.

[0077] Here, "key feature state" refers to the form in which the optimized supplementary data evolves from its initial generation to its final formation, enabling the optimal edge computing nodes to continue performing deeper edge computing on the target data. For example, it might be the initial generation of data comparison results. "Network environment information of the path that should be generated" refers to the map environment information along the path that should be generated in the capability mining scheduling network. This reflects the data collection status of the capability nodes when collecting information about the path that should be generated. Based on this data collection status, it can be determined at which path point the capability nodes will generate the key feature state, and the local path after that path point is then taken as the target local path.

[0078] S5003. Obtain the path segments of multiple acquisition capability nodes that will pass through the target local path in the future, and the changes in acquisition capability during the passage.

[0079] Based on the future information collection paths of each data collection capability node, the path segments of multiple data collection capability nodes that will pass through the target local path in the future (for the limitation of "in the future", the future time length can be preset, and this time length will be considered as "in the future") and the changes in their data collection capabilities during the path can be determined. The changes in data collection capabilities during the path refer to the changes in the data collection capabilities that the data collection capability nodes possess when collecting information along the path segments.

[0080] S5004. Based on the path segments of each data acquisition node and the changes in data acquisition capabilities during the path, explore the relay data acquisition capabilities of each data acquisition node to continuously attempt to acquire specific content of the optimized supplementary data.

[0081] S5005. Based on relay acquisition capability, the specific content of scheduling each acquisition capability node to continuously attempt to acquire optimized supplementary data.

[0082] Specifically, based on the path segments of each data acquisition node and the changes in acquisition capabilities during the path, the relay acquisition capability of each data acquisition node is explored to continuously attempt to collect specific content of the optimized supplementary data. Finally, based on the relay acquisition capability, each data acquisition node is scheduled to continuously attempt to collect specific content of the optimized supplementary data, so as to give full play to its relay acquisition capability.

[0083] The aforementioned technical solution introduces the expected generation path of optimized supplementary data. Based on its network environment information, it determines the target local path along which the optimized supplementary data begins to generate key feature states and develops. Relay data collection capability mining is performed only on the target local path, improving mining efficiency and reducing mining resources. Next, it obtains the path segments of multiple data collection nodes that will pass through the target local path and the changes in their collection capabilities during this process. Based on this, it mines the relay data collection capability of each data collection node to continuously attempt to collect specific content from the optimized supplementary data, and schedules corresponding continuous relay attempts based on this capability. This significantly improves the ability to maximize the collection of specific content from the optimized supplementary data. Notably, this enhanced capability complements the iterative or autonomous reinforcement learning of the edge computing capabilities of edge computing nodes, enabling the system to form a closed loop of capability growth and utilization. This greatly enhances the system's intelligence level and further effectively integrates edge computing technology into the data management system of the entire sorghum industry chain.

[0084] Example 9: In this embodiment of the invention, the relay acquisition capability, which involves mining the relay acquisition capability of each acquisition capability node to continuously attempt to acquire specific content from the optimized supplementary data based on the path segments of each acquisition capability node and the changes in acquisition capability during the path, includes: Each data acquisition node is divided into multiple node sets; in the same node set, the path segments of multiple data acquisition nodes overlap in pairs.

[0085] The phrase "the passage segments overlap in pairs" means, for example, that the passage segments that should be in the order of the path are passage segment 1, passage segment 2, and passage segment 3. There is an overlap between passage segment 1 and passage segment 2, and there is also an overlap between passage segment 2 and passage segment 3. Passage segment 1 and passage segment 3 may or may not overlap directly.

[0086] Iterate through each node set in turn; Each time a node is traversed, the relay collection capacity requirement is reflected by the sum of the path segments of each collection capacity node in the traversed node set. To address the need for relay data collection capabilities, we attempted to identify potential relay data collection capabilities by exploring the changes in data collection capabilities of each data collection capability node during the traversal of the node set. After traversing each node set in sequence, the candidate relay collection capabilities that were successfully mined during each traversal are reduced and optimized to obtain relay collection capabilities.

[0087] In this method, data collection nodes whose respective paths overlap in pairs are included in the same node set, so that data collection nodes in the same node set can take turns collecting information on the sum of their respective paths.

[0088] By traversing each node set in turn, the sum of the path segments of each data collection capability node in the traversed node set will reflect the relay data collection capability requirements, that is, how to collect the relevant generated information on the sum of the path segments.

[0089] To address the need for relay data collection capabilities, we attempt to identify potential relay data collection capabilities among the data collection capabilities of each data collection node in the traversed node set as their data collection capabilities change along their respective paths. These potential relay data collection capabilities can meet the relay data collection capability requirements. Furthermore, they enable each data collection node in the traversed node set to relay and utilize the information collection capabilities that change along their respective paths.

[0090] After traversing each node set in sequence, the candidate relay collection capabilities that were successfully mined during each traversal are reduced and optimized to obtain relay collection capabilities.

[0091] The specific algorithm for the reduction optimization is as follows: Construct a task scheduling table on the timeline. For adjacent data collection nodes A and B in the same node set, let the end time of data collection by node A on its path segment be... The start time for node B to collect data on its path segment is... : Calculate the degree of relay incoordination (Gap): .like This indicates a data collection gap, requiring node rescheduling. Extend working hours or nodes Early intervention. This includes all time variables ( All times are measured in seconds (s) of the system's unified clock or in UNIX timestamp format.

[0092] Calculate the degree of overlap in work: This formula is used to quantify the time overlap between node A and node B, where node A has not yet finished collecting data. If... This indicates that redundant data collection exists.

[0093] Optimize execution: In Within the time period, compare nodes and By optimizing energy consumption costs and eliminating data collection instructions with higher costs, the optimization goal of achieving zero work repetition and zero relay incoordination can be achieved. The energy consumption cost... The calculation formula is: .in, The average power during data acquisition at the acquisition node. For the duration of data collection; The average power is collected when the node moves. The system prioritizes reserving data due to the time spent on the move. Acquisition commands with smaller values.

[0094] Specifically, the steps for reducing and optimizing the candidate relay collection capabilities successfully discovered during each traversal include: taking the degree of work repetition (the degree of repeated information collection) of each collection capability node when scheduling each collection capability node to continuously attempt to collect the specific content of the optimized supplementary data based on the reduced and optimized relay collection capabilities as the reduction and optimization target, and taking the degree of relay incoordination (which can be counted as the average time interval between multiple groups of collection capability nodes completing relay information collection, the larger the time interval, the greater the degree of relay incoordination) as the reduction and optimization target, and performing reduction and optimization on the candidate relay collection capabilities successfully discovered during each traversal to be closest to the reduction and optimization target.

[0095] The aforementioned technical solution divides each data acquisition node into multiple node sets, enabling data acquisition nodes within the same set to relay the collection of information from the sum of their respective path segments, thus improving the efficiency of relay data acquisition capability mining. Secondly, it sequentially traverses each node set, analyzing the relay data acquisition capability requirements reflected by the sum of the path segments of each data acquisition node within the traversed node set. Based on these requirements, it attempts to mine potential relay data acquisition capabilities from the changes in acquisition capabilities during the path journey of each data acquisition node within the traversed node set, improving the accuracy of relay data acquisition capability mining. Next, after sequentially traversing each node set, it reduces and optimizes the potential relay data acquisition capabilities successfully mined during each traversal, obtaining the final relay data acquisition capabilities. This significantly improves the quality of relay data acquisition capability mining and enhances its ability to be used to schedule continuous relay attempts by data acquisition nodes to collect optimized supplementary data.

[0096] Example 10: This invention provides a data management method for the entire sorghum industry chain, such as... Figure 2 As shown, it includes: Feature analysis module 100 is used to analyze the multimodal fuzzy features of target data of the entire sorghum industry chain collected through the Internet of Things; The node matching module 200 is used to match the optimal edge computing node for the target data based on multimodal fuzzy features; The edge computing module 300 is used to transmit the target data to the optimal edge computing node for edge computing, obtain the edge computing results, and output them.

[0097] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A data management method for the entire sorghum industry chain, characterized in that, include: Analyze the multimodal fuzzy features of target data from the entire sorghum industry chain collected through the Internet of Things; Based on multimodal fuzzy features, the optimal edge computing nodes are matched for the target data; The target data is transmitted to the optimal edge computing node for edge computing, and the edge computing results are obtained and output.

2. The sorghum whole-industry chain data management method as described in claim 1, characterized in that, The multimodal fuzzy features include at least: data type, data source type, and spatiotemporal information.

3. The sorghum whole-industry chain data management method as described in claim 1, characterized in that, The method of matching the optimal edge computing node for target data based on multimodal fuzzy features includes: Predict the edge computing needs of target data reflected by multimodal fuzzy features; wherein the types of edge computing needs include at least: sorghum growth monitoring, planting decision support, processing quality control, and product sales management; The candidate edge computing node whose currently available edge computing capacity best meets the edge computing requirements is selected as the optimal edge computing node and matched with the target data.

4. The sorghum whole-industry chain data management method as described in claim 1, characterized in that, When transmitting the target data to the optimal edge computing node for edge computing, if the optimal edge computing node reports that additional supplementary data is needed to support the continued edge computing of the target data, and the supplementary data exceeds the range of IoT collection capabilities, the supplementary data is optimized to achieve an optimal balance between the difficulty of collecting the optimized supplementary data and the support strength of the optimal edge computing node for the continued edge computing of the target data. Based on the data acquisition capabilities, the scheduling network is used to supplement the specific content of the optimized supplementary data to the maximum extent. The additional data collected will be transmitted as feedback to the optimal edge computing node.

5. The sorghum whole-industry chain data management method as described in claim 4, characterized in that, The optimization of the supplementary data aims to achieve an optimal balance between the difficulty of collecting the optimized supplementary data and the support strength of the optimal edge computing node for further edge computing on the target data, including: Plan an N-level optimization scheme to supplement the data; Take i from 1 to N in sequence, and determine the value m of i when the difficulty of collecting the supplementary data after optimization by the i-th level optimization scheme and the support strength of the edge computing node for the target data to continue edge computing are closest to the balance target. The supplementary data optimized by the m-level optimization scheme is selected as the final optimized supplementary data. The balance objectives include: The difficulty of data collection is set at the first equilibrium threshold. The level of support is the second equilibrium threshold.

6. The sorghum whole-industry chain data management method as described in claim 5, characterized in that, The N-level optimization scheme for the supplementary planning data includes: Based on the data type distribution of the supplementary data, suitable optimization scheme planning rules for optimizing the supplementary data are determined from the optimization scheme planning rule base; Based on the optimization scheme planning rules, plan an N-level optimization scheme for supplementary data.

7. The sorghum whole-industry chain data management method as described in claim 5, characterized in that, The steps for obtaining the first balance threshold and the second balance threshold include: The analysis supplements the upper limit of the collection difficulty represented by the collection capability; The lower limit of the support level represented by the degree of supplementary data required to analyze the optimal edge computing node; Based on the relationship between the upper limit of difficulty and the lower limit of intensity, a balance rule is determined from the balance rule base to achieve the optimal balance between the collection difficulty and the collection intensity. The balance rule includes: the upper limit adjustment coefficient and the lower limit adjustment coefficient. Based on the upper limit reduction coefficient, the upper limit of difficulty is reduced to obtain the first balance threshold; Based on the lower limit adjustment coefficient, the lower limit of intensity is adjusted upward to obtain the second equilibrium threshold.

8. The sorghum whole-industry chain data management method as described in claim 4, characterized in that, The specific content of the supplementary data after collection optimization, based on the data acquisition capability mining and scheduling network, includes: In the data acquisition capability mining and scheduling network, the path that should be generated for the optimized supplementary data is represented; Based on the network environment information of the path that should be generated, the target local path that should be generated is determined to start generating key feature states and develops on the optimized supplementary data on the path that should be generated. Acquire the path segments of multiple data acquisition nodes that will pass through the target local path in the future, and the changes in their acquisition capabilities as they pass through the target. Based on the path segments of each data acquisition node and the changes in data acquisition capabilities during the path, we explore the relay data acquisition capabilities of each data acquisition node to continuously attempt to collect specific content from the optimized supplementary data. Based on relay data acquisition capabilities, each acquisition node is scheduled to continuously attempt to acquire the specific content of the optimized supplementary data.

9. The sorghum whole-industry chain data management method as described in claim 8, characterized in that, The relay acquisition capability, which involves continuously attempting to collect specific content from the optimized supplementary data based on the path segments of each acquisition node and the changes in acquisition capability during the path, includes: Each data acquisition node is divided into multiple node sets; in which, the path segments of multiple data acquisition nodes in the same node set overlap in pairs. Iterate through each node set in turn; Each time a node is traversed, the relay collection capacity requirement is reflected by the sum of the path segments of each collection capacity node in the traversed node set. To address the need for relay data collection capabilities, we attempted to identify potential relay data collection capabilities by exploring the changes in data collection capabilities of each data collection capability node during the traversal of the node set. After traversing each node set in sequence, the candidate relay collection capabilities that were successfully mined during each traversal are reduced and optimized to obtain relay collection capabilities.

10. A data management system for the entire sorghum industry chain, characterized in that, include: The feature analysis module is used to analyze the multimodal fuzzy features of target data from the entire sorghum industry chain collected through the Internet of Things; The node matching module is used to match the optimal edge computing node for the target data based on multimodal fuzzy features; The edge computing module is used to transmit target data to the optimal edge computing node for edge computing, obtain edge computing results, and output them.

Images