A big data-based crop full life cycle traceability system

Abstract

The application discloses a kind of crop full life cycle traceability systems based on big data, it is related to intelligent agricultural technology field, including: edge perception computing unit, cloud service platform and alliance block chain network;Edge unit collects multi-modal data, determines phenology stage, selects optimal AI model to analyze and identify agricultural behavior, triggers compliance verification, generates structured traceability event data and is associated with hash value and chain;Cloud storage phenology knowledge graph and SOP rule base;Block chain solidifies operation rules with smart contract, realizes crop full life cycle credible traceability.The application realizes adaptive accurate monitoring and resource optimization by depth fusion and innovative design of edge computing, artificial intelligence, block chain and multi-modal sensing technology, improves traceability information from behavior record to compliance proof, and gives traceability data objective quantitative quality connotation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent agriculture, and particularly relates to a crop full life cycle traceability system based on big data. BACKGROUND

[0002] With the increasing attention of consumers to the quality and safety of agricultural products, and the growing demand of enterprises for standardized management of agricultural product production, the full life cycle traceability technology of crop production has become a research hotspot in the field of intelligent agriculture. Tea, as an important economic crop, its production process traceability is particularly important. The existing tea traceability technology usually combines technologies such as Internet of Things, artificial intelligence and blockchain to realize the collection of tea production environment data, pest and disease identification, and information recording and traceability.

[0003] However, the existing technology still has many defects. The AI recognition model used at present is often general, which cannot be adjusted according to the morphological characteristics of different phenological stages of tea and the susceptible pest and disease species, resulting in poor recognition pertinence, resource waste and low efficiency. The monitoring of agricultural behavior can only record the behavior itself, and cannot judge whether it meets the production standard, so that the traceability information lacks deep verification of the quality of the production process. In the complex environment of tea garden, the lightweight model on the edge computing device has poor robustness and large fluctuation in recognition accuracy, and uploading the task to the cloud will cause network congestion and high delay. In addition, although all raw data are chained, it can guarantee the credibility, but it will bring huge storage and transaction pressure to the blockchain, which is high in cost and impractical. SUMMARY

[0004] The purpose of the present application is to provide a crop full life cycle traceability system based on big data, which solves the problems in the background art.

[0005] To solve the above technical problems, the present application provides a crop full life cycle traceability system based on big data, comprising: an edge perception computing unit for collecting multi-modal data of target crops; the multi-modal data includes current environmental parameters, historical meteorological data and tea tree image data;

[0006] a cloud service platform for storing a preset crop phenological knowledge graph and a preset standard operation procedure rule library;

[0007] a consortium blockchain network for solidifying operation rules from the standard operation procedure rule library in the form of a smart contract;

[0008] The edge perception computing unit is further used for:

[0009] determining the current phenological stage of the target crop based on the collected current environmental parameters and historical meteorological data, and querying the crop phenological knowledge graph.

[0010] Based on the determined current phenology stage, the optimal artificial intelligence model is selected and loaded from the local model library through a model applicability scoring method;

[0011] The collected tea tree image data is analyzed online using the loaded optimal artificial intelligence model to identify agricultural behaviors;

[0012] When the agricultural behavior is identified, compliance verification of the agricultural behavior is triggered;

[0013] The compliance verification is to call the smart contract in the alliance blockchain network to verify the operator identity, operation time, associated materials and geographic location of the agricultural behavior in real time to generate a compliance verification result;

[0014] Also used for generating structured traceability event data based on the compliance verification result;

[0015] The structured traceability event data is associated with the hash value of the multi-modal data and recorded in the alliance blockchain network.

[0016] Preferably, the specific implementation of the model applicability scoring method comprises:

[0017] Determine the preset phenology correlation coefficient;

[0018] Determine the preset model normalized resource consumption;

[0019] Multiply the phenology correlation coefficient by the preset first weight coefficient to obtain a phenology correlation weighted value;

[0020] Multiply the model normalized resource consumption by the preset second weight coefficient to obtain a resource consumption weighted value;

[0021] Subtract the resource consumption weighted value from the phenology correlation weighted value to generate a model applicability score for deciding which artificial intelligence model to load.

[0022] Preferably,

[0023] The phenology correlation coefficient is a preset numerical value for representing the importance of each artificial intelligence model at different phenology stages;

[0024] The preset numerical value is predefined according to agronomic expert knowledge for each phenology stage;

[0025] The model normalized resource consumption is a numerical value obtained by offline benchmarking of each artificial intelligence model on a target edge hardware and normalizing the test results;

[0026] The first weight coefficient and the second weight coefficient are preset adjustable parameters according to business requirements.

[0027] Preferably, the specific implementation of the compliance verification includes:

[0028] An operator verification function in the smart contract is called to determine whether the operator identity is in a preset authorized personnel list.

[0029] A time verification function in the smart contract is called to determine whether the operation time is within a preset compliance time window.

[0030] A material verification function in the smart contract is called to determine whether the associated material is within a preset allowed use list.

[0031] The compliance verification result is true only when all verification functions return positive results; otherwise, the compliance verification result is false.

[0032] Preferably, the multi-modal data collected by the edge-aware computing unit also includes spectral data.

[0033] Preferably, it further includes a tea quality online evaluation module for extracting a plurality of original quality characteristic values based on the loaded optimal artificial intelligence model and the collected spectral data.

[0034] The original quality characteristic values include bud leaf ratio representing tea bud morphology, color uniformity representing leaf color, and spectral absorption peak representing tea content.

[0035] Each of the original quality characteristic values is normalized and multiplied by a respective quality weight coefficient, and all products are summed to generate a tea quality evaluation score.

[0036] The structured traceability event data is also generated based on the tea quality evaluation score.

[0037] Preferably, the quality weight coefficient is obtained by a machine learning regression modeling method, and the regression modeling method includes the following specific steps: collecting multi-modal feature data of a large number of tea samples and sensory evaluation true scores given by tea evaluation experts, taking the multi-modal feature data as the independent variable and the sensory evaluation true scores as the dependent variable, training a multiple linear regression model, and taking the regression coefficients of the normalized multiple linear regression model as the quality weight coefficient.

[0038] Preferably, it is also used to trigger an alarm and record the agricultural behavior that leads to false verification as an abnormal event when the compliance verification result is false.

[0039] Compared with the prior art, the present application has the following beneficial effects:

[0040] Through the phenology perception mechanism, the optimal artificial intelligence model can be autonomously and intelligently selected and loaded according to the biological rhythm of the crops, thereby improving the pertinence and accuracy of identification and optimizing the resource utilization rate of the edge device. Through the data such as the operator identity, operation time and associated materials captured in real time, the SOP intelligent cross verification is introduced, and the rules are solidified by using the blockchain smart contract, so that the traceability information is upgraded from simple behavior records to compliance proof, thereby enhancing the value and public credibility of the traceability information. 3. By combining the biological cycle of crops, production standards and real-time monitoring data, a digital twin is constructed, which provides a data basis for precise management, yield prediction and quality optimization, and improves the data comprehensive utilization capability. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0042] Fig. 1 It is a logical block diagram of the system of the present application;

[0043] Fig. 2 It is a specific implementation step diagram of the model applicability scoring method of the present application;

[0044] Fig. 3 It is a specific implementation step diagram of the compliance verification of the present application. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0046] Embodiment 1:

[0047] Please refer to Figs. 1-3The application provides a crop full life cycle traceability system based on big data, comprising: an edge perception computing unit for collecting multi-modal data of target crops; the multi-modal data comprises current environmental parameters, historical meteorological data and tea tree image data; a cloud service platform for storing a preset crop phenology knowledge graph and a preset standard operation procedure rule library; a consortium blockchain network for solidifying operation rules from the standard operation procedure rule library in the form of a smart contract; the edge perception computing unit is further used for: determining the current phenology stage of the target crop based on the collected current environmental parameters and the historical meteorological data and querying the crop phenology knowledge graph; selecting and loading an optimal artificial intelligence model from a local model library based on the determined current phenology stage through a model applicability scoring method; performing online analysis on the collected tea tree image data by using the loaded optimal artificial intelligence model to identify agricultural activities; when the agricultural activities are identified, triggering compliance verification of the agricultural activities; the compliance verification is to call the smart contract in the consortium blockchain network to verify the operator identity, operation time, associated materials and geographic location of the agricultural activities in real time to generate a compliance verification result; the system is further used for generating structured traceability event data based on the compliance verification result; associating the structured traceability event data with the hash value of the multi-modal data and recording them in the consortium blockchain network.

[0048] A crop full life cycle traceability system based on big data, which builds a full-link trusted closed loop from perception, analysis to recording; the edge perception computing unit deployed in the tea garden site is used for continuously collecting multi-modal data, which comprehensively describes the tea tree growth environment, specifically including current environmental parameters such as real-time air temperature, humidity and light intensity obtained through environmental sensors, combined with historical meteorological data synchronized from the cloud service platform, and tea tree image data captured through a high-definition camera; at the same time, the cloud service platform as the central knowledge base and rule base of the system, pre-stores the tea phenology knowledge graph constructed by agronomists and the standard operation procedure SOP rule library formulated by enterprises; the consortium blockchain network carries out the task of solidifying the core rules in the SOP rule library in the form of a smart contract, which ensures the non-tamperability and execution transparency of the operation specification;

[0049] The operation logic of the system starts from the edge perception computing unit, integrates the collected current environmental parameters and historical meteorological data, and queries and compares with the phenology knowledge graph to accurately determine the specific phenology stage of the current tea tree, such as the sprouting period or the picking period. This determination constitutes the basis for all subsequent intelligent analysis. Instead of using a static and fixed analysis model, after determining the phenology stage, the system immediately starts a model applicability scoring method, dynamically selects and loads an optimal model from multiple candidate artificial intelligence models stored locally based on the monitoring priority of the current stage. When the optimal artificial intelligence model loaded, such as an image analysis model for tender bud recognition, identifies the picking agricultural operation from the real-time image data stream, the system will immediately trigger a compliance verification process. This verification process is not completed locally or in a centralized cloud, but by calling a smart contract deployed on the alliance blockchain network to strictly verify the operator's identity, current operation time, materials used, and geographic location of the operation in real time. After verification, the system generates a structured traceability event data containing verification results, timestamps, operator information, and other elements. To ensure the integrity and non-repudiation of the original data, the system further associates the hash value of the original multi-modal data with the structured traceability event data using cryptography. This complete record containing event details and data fingerprints is finally written into the alliance blockchain network, forming a highly credible, traceable, and information-rich traceability chain, and achieving deep, dynamic, and credible monitoring of the entire life cycle of tea production.

[0050] Embodiment 2:

[0051] The specific implementation of the model applicability scoring method includes: determining a preset phenology correlation coefficient; determining a preset model normalized resource consumption; multiplying the phenology correlation coefficient by a preset first weight coefficient to obtain a phenology correlation weighted value; multiplying the model normalized resource consumption by a preset second weight coefficient to obtain a resource consumption weighted value; subtracting the resource consumption weighted value from the phenology correlation weighted value to generate a model applicability score for deciding which artificial intelligence model to load.

[0052] The phenology correlation coefficient is a preset value used to represent the importance of each artificial intelligence model in different phenology stages; the preset value is predefined according to agronomic expert knowledge for each phenology stage; the model normalized resource consumption is a value obtained by offline benchmark testing of each artificial intelligence model on the target edge hardware and normalizing the test results; the first weight coefficient and the second weight coefficient are adjustable parameters preset according to business requirements;

[0053] The essence of the model applicability scoring method is to establish a quantitative decision mechanism to intelligently schedule the most suitable artificial intelligence model at a specific phenological stage; in the traditional scheme, the model is fixed after deployment, and this static deployment ignores the dynamic changes of monitoring focus and morphological characteristics of crops in different biological cycles; to solve the problem of weak pertinence and resource waste caused by static deployment, the scoring method is designed to quantitatively integrate expert knowledge in the field of agronomy and the constraints of computing resources of edge devices, thereby realizing the intelligentization and efficiency of model scheduling;

[0054] The quantitative implementation of the score relies on a model applicability score formula:

[0055] ;

[0056] The model applicability score is a dimensionless real number, which is the direct basis for the final decision, and the higher the score, the more suitable the model;

[0057] The model index is a positive integer that uniquely identifies the ith candidate artificial intelligence model in the local model library;

[0058] The phenological stage index is a positive integer that uniquely identifies the current growth stage of tea;

[0059] The ith candidate artificial intelligence model is represented by

[0060] The jth current phenological stage is represented by the system according to environmental data and phenological knowledge graph query;

[0061] The phenological relevance coefficient is a dimensionless value between 0 and 1, representing the importance or necessity of running the model at the phenological stage ;

[0062] The model normalized resource consumption is a dimensionless value between 0 and 1, reflecting the computational cost of the model when running on the target edge hardware;

[0063] And are the first weight coefficient and the second weight coefficient, respectively, which are positive real numbers, and are preset adjustable parameters according to business needs, representing the importance of phenological relevance and resource consumption in decision-making, respectively, and are usually set to 1, i.e. , so as to facilitate adjustment;

[0064] The pre-configuration phase of the system involves the rigorous definition of various parameters; the determination of the phenology correlation coefficient depends on the domain knowledge of agronomists; the expert team will evaluate the importance of each phenological stage (such as the germination period, leaf expansion period) and each monitoring task (corresponding to the artificial intelligence model) in that stage and assign a value between 0 and 1. These correspondences are structured and stored in the phenology knowledge graph in the cloud; the determination of the normalized resource consumption of the model is completed through rigorous offline benchmarking; before model deployment, each candidate artificial intelligence model is actually run on the target edge hardware, and performance profiling tools are used to measure its key resource consumption indicators such as floating point operations per second or average power consumption. After collecting all the test data of the models, the original consumption values are mapped to the [0, 1] interval using the min-max normalization method; this normalization process follows the following formula:

[0065] ;

[0066] represents the original resource consumption value (such as average power consumption or floating point operations per second) obtained by offline benchmarking of the candidate model ;

[0067] represents the minimum value of the original resource consumption values of all candidate models in the model library;

[0068] represents the maximum value of the original resource consumption values of all candidate models in the model library;

[0069] Through this formula, the resource consumption of all models can be fairly mapped to 0 to 1 for comparison;

[0070] The first weight coefficient and the second weight coefficient are set by the system administrator according to specific business strategies, for example, in the scenario of pursuing extreme tea quality, we can set ; in the scenario of limited power supply of edge devices, we can set ;

[0071] In a specific application scenario, when the system determines that the current phenological stage is the spring germination period , and there are aphid identification models and tender bud morphology grading models in the local model library; according to the pre-set expert knowledge base, , ; after offline benchmarking, , ; if the current business policy prioritizes tea quality, the weight is set as ; at this time, the score of the model is ; the score of the model is ; because , the system will load and run the tender bud morphology grading model with a higher score ; the introduction of this decision mechanism ensures that the system can dynamically invest limited edge computing resources into the current most valuable monitoring task, significantly improving the relevance of monitoring and resource utilization efficiency.

[0072] Embodiment 3:

[0073] The specific implementation mode of the compliance verification includes: calling an operator verification function in the smart contract to determine whether the operator identity is in a preset authorized personnel list; calling a time verification function in the smart contract to determine whether the operation time is in a preset compliance time window; calling a material verification function in the smart contract to determine whether the associated material is in a preset allowed use list; and determining that the compliance verification result is true only when all verification functions return positive results; otherwise, determining that the compliance verification result is false.

[0074] The system is also used to trigger an alarm and record the farming behavior that leads to a false verification as an abnormal event when the compliance verification result is false.

[0075] The implementation of compliance verification is technically essential to convert the production management standard SOP from a static document into an automatically executable and tamper-proof on-chain code, thereby providing a deep compliance proof for the traceability information. Traditional traceability systems only record behaviors themselves, but cannot determine whether they are compliant, resulting in limited information value. To solve this problem, the present application solidifies the SOP rules in the smart contract of the alliance blockchain, realizes the automatic and real-time audit of on-site farming operations by designing a set of logically rigorous verification functions, and upgrades the traceability information from "behavior record" to "compliance proof".

[0076] From a logical level, compliance verification is a Boolean AND operation composed of multiple sub-verification functions, and its result is jointly determined by the logic deployed in the smart contract, which can be expressed by the following logical formula:

[0077] ;

[0078] represents the final SOP compliance verification result, which is a Boolean value of true (1) or false (0);

[0079] is a logical AND operator;

[0080] represent sub-verification functions for operator verification, time verification, material verification, and geographical location verification, which are pre-encoded in the smart contract, and their internal judgment rules are derived from the SOP file;

[0081] represents the operator identity, which can be derived from the output of a face recognition model or the unique ID of an RFID badge;

[0082] represents the current operation timestamp, which is provided by the system clock calibrated by the edge device via the NTP protocol, in the format of Unix timestamp;

[0083] represents the associated material identification, which is usually derived from the unique ID of the pesticide bottle or packaging features recognized by a visual model;

[0084] represents the geographical location coordinates, which are read in real-time by the device's built-in GPS module;

[0085] When the system detects a pesticide spraying behavior, the edge device immediately captures the on-site operator identity , current operation timestamp , associated material identification , and geographical location coordinates ; the system then calls the function in the smart contract and inputs these parameters; the logic inside the smart contract starts to execute: function checks whether exists in the pre-set list of authorized agricultural technicians; function checks whether is outside the "15-day safety interval before the picking period" specified in the SOP; function checks whether is within the list of allowed organic pesticide materials; function verifies whether is within the specified tea garden plot range;

[0086] Only when all sub-verification functions return true, the final result of is true, and the system determines that this pesticide spraying behavior is compliant and records it as a compliant event on the chain; otherwise, if any of the verifications fails, for example, if an unlisted pesticide is used, resulting in returning false, The result is false; at this time, the system will immediately trigger the alarm mechanism, send an alarm to the management background, and record the violation together with the specific failed verification item as an abnormal event on the blockchain; through this verification and recording mechanism, the system not only realizes the automatic rigid constraint of the production process and effectively prevents illegal operations, but more importantly, it gives each traceable event on the chain irrefutable credibility and deep value, providing a solid technical endorsement for high-end tea brands.

[0087] Embodiment 4:

[0088] The multi-modal data collected by the edge-aware computing unit also includes spectral data;

[0089] It also includes a tea quality online evaluation module for extracting a plurality of original quality characteristic values based on the loaded optimal artificial intelligence model and the collected spectral data; the original quality characteristic values include: bud leaf ratio representing tea bud morphology, color uniformity representing leaf color, and spectral absorption peak representing tea content; each original quality characteristic value is normalized and multiplied by its corresponding quality weight coefficient, and then all the products are summed to generate a tea quality evaluation score; the structured traceability event data is also generated based on the tea quality evaluation score;

[0090] The quality weight coefficient is obtained by a machine learning regression modeling method, and the regression modeling method includes the following specific steps: collecting a large number of tea sample multi-modal feature data and sensory evaluation true scores given by tea evaluation experts, and taking the multi-modal feature data as the independent variable and the sensory evaluation true score as the dependent variable, training a multiple linear regression model, and the regression coefficient of the normalized multiple linear regression model is used as the quality weight coefficient;

[0091] In order to integrate the core value dimension of product quality into the traceability system, the system integrates a tea quality online evaluation module; traditional traceability systems pay more attention to the safety of the production process, and less to the quality dimension of the final product; the value of tea is highly related to its morphology, color, and content, etc. quality indicators, but the traditional evaluation method of these indicators relies on expert sensory evaluation, which has strong subjectivity, cannot be scaled, and is difficult to trace; the present application proposes an online evaluation method that integrates multi-modal data, which establishes an objective and quantitative quality evaluation model through artificial intelligence model and spectral analysis, and converts high-dimensional sensory quality into recordable and verifiable digital scores;

[0092] The quantitative evaluation of tea quality is completed through a quality score fusion formula:

[0093] ;

[0094] is the tea leaf comprehensive quality score, which is a final calculated real number that can be mapped to an interval such as 0-100;

[0095] is the feature index, a positive integer from 1 to N;

[0096] is the total number of original quality feature values for evaluation, the specific number and category of which are determined by the domain knowledge of tea leaf evaluation experts;

[0097] represents the kth original quality feature value, which comes from the direct output of various sensors and artificial intelligence models, for example, the proportion of buds and leaves from image analysis model output, the absorption peak value of a spectrum sensor at a specific waveband to represent the content of a specific tea leaf ingredient;

[0098] is a standard normalization function, such as min-max normalization, used to map original feature values of different dimensions to a unified [0, 1] interval to eliminate the effect of dimensions;

[0099] represents the quality weight coefficient of the kth feature, which represents the contribution of the feature to the final quality score, and all weight coefficients are positive real numbers and their sum ;

[0100] The specific calculation formula is:

[0101] ;

[0102] is the measured value of the kth original quality feature value;

[0103] and are the minimum and maximum values of the reasonable value range of the feature value determined according to a large amount of historical sample data or expert knowledge, respectively;

[0104] In tea leaf picking operations, when the edge perception computing unit collects image and spectrum data of a specific batch of tea leaves, the tea leaf quality online evaluation module is immediately activated; assuming that the system is configured with two quality features (N=2): the bud integrity ( raw value) is obtained through AI image analysis, and the spectrum absorption peak value ( raw value) related to a certain ingredient is obtained through spectrum analysis; after normalization, we get and ; the quality weight coefficient obtained by machine learning regression modeling is and ; the quality score of the batch of tea is ; the system can multiply it by 100 to obtain a final score of 82 points; this quality evaluation score will be combined with the compliance verification result of the batch of picking, operator information, etc. to form a structured traceability event data, and recorded on the blockchain; this design enables consumers to not only verify the compliance of the production process when tracing the product, but also obtain the quality score objectively evaluated at the time of picking of the batch of tea, thereby constructing a digital twin that reflects the true state of the tea garden, and providing a solid data foundation for the precise management, yield prediction and quality optimization of tea;

[0105] The determination of the quality weight coefficient is not subjective, but is obtained by a data-driven machine learning regression modeling method; this method ensures the objectivity and scientificity of the weight coefficient, and the core steps are as follows: a large number of representative tea samples are collected; for each sample, on the one hand, the edge perception calculation unit of the system collects its multi-modal feature data and extracts a plurality of original quality characteristic values ; on the other hand, a plurality of experienced tea evaluation experts are organized to conduct standard sensory evaluation on these samples and give an authoritative true score; then, the collected multi-modal feature data is used as the independent variable (feature matrix) of the model, and the sensory evaluation true score given by the experts is used as the dependent variable (target vector) to train a multiple linear regression model; after the model is trained, the regression coefficients of each independent variable learned by the model objectively reflect the influence degree of each characteristic on the final sensory score; finally, these regression coefficients are normalized, for example, by dividing by the sum of the absolute values of all coefficients to make their sum equal to 1, and the result obtained is the quality weight coefficient ;

[0106] This normalization process is to convert the original regression coefficients (which may be positive or negative and have different dimensions) into dimensionless weights that represent their contribution to the final quality score; the specific calculation formula is as follows:

[0107] ;

[0108] wherein is the original regression coefficient corresponding to the kth quality characteristic obtained after training of the multiple linear regression model, is the absolute value thereof, is the total number of quality characteristics. The use of absolute values ensures that all weights are positive, and the sum of all weights strictly equal to 1, which makes it can be clearly explained as the proportion of the importance of the kth feature in the quality comprehensive evaluation.

[0109] The above is only the preferred embodiment of the present application, and does not limit the present application in other forms. Any skilled person in the art can use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes applied to other fields, but without departing from the technical solution content of the present application. Any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application still belongs to the protection scope of the technical solution of the present application.

Claims

1. A big data based crop life cycle traceability system, characterized in that, The method comprises the following steps: An edge perception computing unit is used to collect multi-modal data of target crops; the multi-modal data comprises current environmental parameters, historical meteorological data, and tea tree image data; A cloud service platform is used to store a preset crop phenology knowledge graph and a preset standard operating procedure rule library; An alliance blockchain network is used to solidify operating rules derived from the standard operating procedure rule library in the form of a smart contract; The edge perception computing unit is further used to: Determine the current phenology stage of the target crop based on the collected current environmental parameters and historical meteorological data, and query the crop phenology knowledge graph; Select and load an optimal artificial intelligence model from a local model library based on the determined current phenology stage through a model applicability scoring method; Perform online analysis on the collected tea tree image data by using the loaded optimal artificial intelligence model to identify agricultural activities; When the agricultural activities are identified, trigger compliance verification of the agricultural activities; The compliance verification is to call the smart contract in the alliance blockchain network to verify the operator identity, operation time, associated materials, and geographic location of the agricultural activities in real time to generate a compliance verification result; Further used to generate structured traceability event data based on the compliance verification result; Associate the structured traceability event data with the hash value of the multi-modal data, and record them together in the alliance blockchain network; The specific implementation of the model applicability scoring method comprises: Determine a preset phenology correlation coefficient; Determine a preset model normalized resource consumption; Multiply the phenology correlation coefficient by a preset first weight coefficient to obtain a phenology correlation weighted value; Multiply the model normalized resource consumption by a preset second weight coefficient to obtain a resource consumption weighted value; Subtract the resource consumption weighted value from the phenology correlation weighted value to generate a model applicability score for deciding which artificial intelligence model to load; The phenology correlation coefficient is a preset value used to represent the importance of each artificial intelligence model at different phenology stages; The preset value is predefined for each phenology stage according to agronomic expert knowledge; The model normalized resource consumption is a value obtained by offline benchmark testing of each artificial intelligence model on a target edge hardware, measuring its key resource consumption indicators using a performance profiling tool, and normalizing the test results; The first weight coefficient and the second weight coefficient are adjustable parameters preset according to business requirements.

2. The big data based crop lifecycle traceability system according to claim 1, wherein, The specific implementation of the compliance verification comprises: Calling an operator verification function in the smart contract to determine whether the operator identity is in a preset list of authorized personnel; Calling a time verification function in the smart contract to determine whether the operation time is within a preset compliance time window; Calling a material verification function in the smart contract to determine whether the associated materials are within a preset list of allowed use. The compliance verification result is determined to be true if and only if all the verification functions return positive results; otherwise, the compliance verification result is determined to be false.

3. The big data based crop lifecycle traceability system according to claim 1, wherein, The multi-modal data collected by the edge-aware computing unit further includes spectral data.

4. The big data based crop lifecycle traceability system according to claim 3, wherein, The tea quality online evaluation module is further included for extracting a plurality of original quality characteristic values based on the loaded optimal artificial intelligence model and the collected spectral data. The original quality characteristic values include: bud leaf ratio representing the shape of tea bud, color uniformity representing the color of tea leaf, and spectral absorption peak value representing the content of tea leaf. Each of the original quality characteristic values is normalized, multiplied by a corresponding quality weight coefficient, and all products are summed to generate a tea quality evaluation score. The structured traceability event data are further generated based on the tea quality evaluation score.

5. The big data based crop life cycle traceability system according to claim 4, wherein, The quality weight coefficient is obtained by a machine learning regression modeling method, which includes the following specific steps: collecting multi-modal feature data of a large number of tea samples and real sensory evaluation scores given by tea evaluation experts, taking the multi-modal feature data as the independent variable and the real sensory evaluation scores as the dependent variable, training a multiple linear regression model, and taking the regression coefficients of the normalized multiple linear regression model as the quality weight coefficients.

6. The big data based crop life cycle traceability system according to claim 2, wherein, The method is further used for triggering an alarm and recording the agricultural behavior leading to the false verification as an abnormal event when the compliance verification result is false.

Images