Intelligent identification and management system and method for key commodities of vegetable basket

Abstract

The invention discloses an intelligent identification and management system and method for key commodities of a vegetable basket, and relates to the technical field of computer processing, and the system comprises a multi-modal data collection and preprocessing module which is used for collecting commodity images, temperature and humidity curves, vibration frequencies and electronic quality inspection reports through Internet of Things equipment, the block chain technology is adopted to encrypt and store production batches and detection values; and the intelligent labeling and label generation module is used for performing instance segmentation on the commodity image, analyzing semantic information of the quality inspection report and generating a multi-level quality label. According to the invention, through a multi-modal data fusion and dynamic association analysis technology, image recognition, sensor data and text analysis are integrated, an intelligent supervision system covering a production-retail full link is constructed, and on the basis of an adversarial training mechanism and an edge cloud collaborative architecture, the production-retail full link is completely supervised. The real-time identification capability of complicated illegal behaviors such as water-injected meat, pesticide residue exceeding and label tampering is obviously improved, and the whole-process safety guarantee of livelihood commodities is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of computer processing technology, and in particular to an intelligent identification and management system and method for key commodities in the vegetable basket. Background Art

[0002] The "vegetable basket" project, a crucial foundation for ensuring people's livelihoods, encompasses the entire production, distribution, and sales chain of agricultural and sideline products, including vegetables, fruits, meat, and aquatic products. In recent years, with rising consumption and heightened food safety requirements, traditional vegetable management practices have become increasingly outdated and insufficiently effective.

[0003] To address these issues, existing technologies rely on manual record-keeping or single identifiers (such as QR codes / RFID), making it difficult to achieve full data connectivity across the supply chain. Fresh produce often changes hands, leading to data fragmentation, and data formats between farmers, logistics providers, and other entities are not interoperable, creating "data silos." For example, Chinese patent publication number CN104834999A discloses a cloud-based method and system for commodity traceability and consumer flow statistical analysis. While capable of tracking commodity flows, it cannot dynamically correlate multimodal data such as images, quality inspection reports, and logistics temperature and humidity. The resulting traceability is coarse-grained, making it difficult to identify violations such as selling inferior goods as genuine goods. Summary of the Invention

[0004] In response to the above technical problems, the technical solution adopted by the present invention is an intelligent identification and management system for key commodities in the vegetable basket, including: A multimodal data acquisition and preprocessing module, which uses IoT devices to collect product images, temperature and humidity curves, vibration frequencies, and electronic quality inspection reports, and uses blockchain technology to encrypt and store production batches and test values; Intelligent annotation and label generation module, used to perform instance segmentation on product images, parse semantic information in quality inspection reports, and generate multi-level quality labels; The multimodal fusion model training module uses a three-stream fusion network to simultaneously train image, text, and time series data, and outputs product quality prediction results; The spatiotemporal correlation traceability engine module establishes spatiotemporal correlation rules for multimodal data and generates product traceability maps; Edge-cloud collaborative monitoring module deploys edge computing nodes to achieve real-time quality monitoring and multi-level early warning; The visual supervision platform module dynamically displays the full-link traceability information and risk evidence chain of the product; The model iteration optimization module optimizes model parameters and traceability rule base based on consumer feedback data.

[0005] Preferably, the multimodal data acquisition and preprocessing module includes: IoT sensor unit, used to collect images, temperature and humidity curves, vibration frequency data and electronic quality inspection reports of circulation nodes; The spatiotemporal alignment unit performs spatiotemporal alignment on multi-source heterogeneous data and cleans invalid segments; The blockchain evidence storage unit encrypts and stores production batches, transportation tracks, and test values.

[0006] Preferably, the multimodal data acquisition and preprocessing module includes: IoT sensor unit, used to collect images, temperature and humidity curves, vibration frequency data and electronic quality inspection reports of circulation nodes; The spatiotemporal alignment unit performs spatiotemporal alignment on multi-source heterogeneous data and cleans invalid segments; The blockchain evidence storage unit encrypts and stores production batches, transportation tracks, and test values.

[0007] Preferably, the intelligent annotation and label generation module includes: The visual annotation unit uses an improved Mask R-CNN model to perform pixel-level segmentation on product images and mark areas of appearance defects and damaged packaging. Semantic parsing unit, which extracts pesticide residue content and freshness score indicators from quality inspection reports based on natural language processing technology; The label association unit associates image features with quality inspection indicators to generate a three-level quality label body including qualified products, critical products, and non-compliant products.

[0008] Preferably, the edge-cloud collaborative monitoring module includes: Edge detection unit, a lightweight model deployed on cold chain vehicles, performs real-time package damage detection and triggers local re-inspection; A multi-level warning unit sets threshold rules for triggering event recording when the temperature deviates by 2°C, and triggering verification of the container door magnetic sensor when the temperature deviates by 5°C; The conflict freezing unit automatically freezes the outbound operation of the batch when the color characteristics of the product image conflict with the quality inspection maturity index.

[0009] Preferably, the spatiotemporal correlation tracing engine module includes: Rule reasoning unit, which establishes spatiotemporal association rules between transport temperature and humidity anomalies and image feature changes; The traceability code generation unit aggregates raw material data from the production end, environmental records from the circulation end, and re-inspection results from the sales end to generate a unique traceability code; The violation mining unit applies graph neural networks to analyze cross-batch data and identify the behavior chain of suppliers replacing origin labels.

[0010] Preferably, the visual supervision platform module includes: A three-dimensional heat map unit shows the spatial correlation between regional complaint rates and transportation temperature and humidity; The intelligent inspection unit automatically associates the supplier's historical violation records with merchants who continuously inspect critical products; The evidence chain encapsulation unit generates tamper-proof regulatory reports based on abnormal image sequences, abnormal sensor data, and quality inspection contradictions.

[0011] Preferably, the model iteration optimization module collects quality evaluation data through the consumer-side feedback interface, uses an incremental learning algorithm to update the multimodal fusion model parameters, and optimizes the traceability rule base logic based on association rule mining.

[0012] A method for intelligently identifying and managing key commodities in the vegetable basket, for implementing the system for intelligently identifying and managing key commodities in the vegetable basket described in the above embodiment, comprises the following steps: S01. Collect multimodal product data, including product images, quality inspection reports, logistics temperature and humidity sensor data, and supplier information, perform data cleaning, standardization, and cloud storage to build a multimodal product database; S02. Based on the multimodal product database, perform pixel-level segmentation and annotation on product images, parse semantic information in quality inspection reports and associate them with sensor data to generate multi-level quality labels; S03. Train a multimodal fusion deep learning model to simultaneously process image features, quality inspection text, and sensor time series data, and output product category, quality grade, and abnormal risk prediction results; S04. Establish spatiotemporal association rules for multimodal data, dynamically match product image features, quality inspection index fluctuations, and logistics environment changes, and generate a fine-grained product traceability map; S05. Deploy an edge-cloud collaborative monitoring mechanism to perform real-time product image comparison, temperature and humidity deviation warning, and quality deterioration prediction; S06. Visualize the traceability information of the entire product chain, the distribution of risk hotspots, and the evidence chain of violations; S07. Iteratively optimize model parameters and traceability rule base logic based on consumer feedback data.

[0013] Preferably, the deployment of the edge-cloud collaborative monitoring mechanism in step S05 includes: S51. Deploy edge computing devices at cold chain transportation nodes and run lightweight detection models to identify package damage in real time. S52. Set multi-level temperature and humidity deviation thresholds: when the temperature deviates by 2°C, an abnormal event is recorded; when it deviates by 5°C, the container door magnetic sensor is activated to verify illegal unpacking behavior; S53. When it is detected that the color characteristics of the product image are inconsistent with the maturity index of the quality inspection report, the batch of products is automatically intercepted from leaving the warehouse and the manual review process is triggered.

[0014] The present invention has at least the following beneficial effects: 1. Through multimodal data fusion, dynamic correlation analysis and full-link monitoring, the problem of the existing traceability system having a single data dimension and difficulty in identifying complex violations can be solved.

[0015] 2. The system integrates image recognition, text analysis, and sensor data processing technologies to build a fine-grained traceability map, enabling intelligent supervision of the entire process from production bases to retail terminals.

[0016] 3. Enhance the model's anti-fraud capabilities through adversarial training mechanisms, and establish an edge-cloud collaborative architecture to ensure real-time response. This can effectively identify illegal operations such as water-injected meat, fruits and vegetables with excessive pesticide residues, and tampered labels, thereby improving the safety of people's livelihood products. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A module structure diagram provided for the first embodiment of the present invention; Figure 2 This is a unit architecture diagram of the multimodal data acquisition and preprocessing module provided in Example 1 of the present invention; Figure 3 This is a unit architecture diagram of the intelligent annotation and label generation module provided in Example 1 of the present invention; Figure 4 This is a unit architecture diagram of the multimodal fusion model training module provided in Example 1 of the present invention; Figure 5 This is a unit architecture diagram of the edge-cloud collaborative monitoring module provided in Example 1 of the present invention; Figure 6 This is a unit architecture diagram of the spatiotemporal correlation tracing engine module provided in Example 1 of the present invention; Figure 7 This is a unit architecture diagram of the visual supervision platform module provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way are interchangeable where appropriate so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0021] Example 1

[0022] This embodiment provides a method and method for intelligently identifying and managing key commodities in the food basket, the method comprising the following steps: Figure 1 As shown: A multimodal data acquisition and preprocessing module, which uses IoT devices to collect product images, temperature and humidity curves, vibration frequencies, and electronic quality inspection reports, and uses blockchain technology to encrypt and store production batches and test values; Specific, combined Figure 2 As shown, the above modules include: IoT sensor unit, used to collect images, temperature and humidity curves, vibration frequency data and electronic quality inspection reports of circulation nodes; The spatiotemporal alignment unit performs spatiotemporal alignment on multi-source heterogeneous data and cleans invalid segments; The blockchain evidence storage unit encrypts and stores production batches, transportation tracks, and test values.

[0023] Intelligent annotation and label generation module, used to perform instance segmentation on product images, parse semantic information in quality inspection reports, and generate multi-level quality labels; Specific, combined Figure 3 As shown, the above modules include: The visual annotation unit uses an improved Mask R-CNN model to perform pixel-level segmentation on product images and mark areas of appearance defects and damaged packaging. Semantic parsing unit, which extracts pesticide residue content and freshness score indicators from quality inspection reports based on natural language processing technology; The label association unit associates image features with quality inspection indicators to generate a three-level quality label body including qualified products, critical products, and non-compliant products.

[0024] The multimodal fusion model training module uses a three-stream fusion network to simultaneously train image, text, and time series data, and outputs product quality prediction results; Specific, combined Figure 4 As shown, the above modules include: The three-stream fusion network architecture consists of a CNN branch, a BiLSTM branch, and a TCN branch to process images, text, and time series data respectively; Dynamic attention unit, which adjusts feature weights through a cross-modal attention mechanism to identify hidden quality degradation caused by cold chain interruptions; The adversarial training unit generates adversarial samples with water injection, weight increase, and expired label tampering to improve the model's anti-interference ability.

[0025] The spatiotemporal correlation traceability engine module establishes spatiotemporal correlation rules for multimodal data and generates product traceability maps; Specific, combined Figure 4 As shown, the above modules include: Rule reasoning unit, which establishes spatiotemporal association rules between transport temperature and humidity anomalies and image feature changes; The traceability code generation unit aggregates raw material data from the production end, environmental records from the circulation end, and re-inspection results from the sales end to generate a unique traceability code; The violation mining unit applies graph neural networks to analyze cross-batch data and identify the behavior chain of suppliers replacing origin labels.

[0026] Edge-cloud collaborative monitoring module deploys edge computing nodes to achieve real-time quality monitoring and multi-level early warning; Specific, combined Figure 5 As shown, the above modules include: Edge detection unit, a lightweight model deployed on cold chain vehicles, performs real-time package damage detection and triggers local re-inspection; A multi-level warning unit sets threshold rules for triggering event recording when the temperature deviates by 2°C, and triggering verification of the container door magnetic sensor when the temperature deviates by 5°C; The conflict freezing unit automatically freezes the outbound operation of the batch when the color characteristics of the product image conflict with the quality inspection maturity index.

[0027] The visual supervision platform module dynamically displays the full-link traceability information and risk evidence chain of the product; Specific, combined Figure 5 As shown, the above modules include: A three-dimensional heat map unit shows the spatial correlation between regional complaint rates and transportation temperature and humidity; The intelligent inspection unit automatically associates the supplier's historical violation records with merchants who continuously inspect critical products; The evidence chain encapsulation unit generates tamper-proof regulatory reports based on abnormal image sequences, abnormal sensor data, and quality inspection contradictions.

[0028] The model iteration optimization module optimizes model parameters and traceability rule base based on consumer feedback data, collects quality evaluation data through the consumer-side feedback interface, uses incremental learning algorithms to update multimodal fusion model parameters, and optimizes the traceability rule base logic based on association rule mining.

[0029] In summary, through multimodal data fusion, dynamic correlation analysis, and full-chain monitoring, the existing traceability system's single data dimension and difficulty in identifying complex violations can be addressed. The system also integrates image recognition, text analysis, and sensor data processing technologies to construct a fine-grained traceability map, enabling intelligent supervision of the entire process from production base to retail terminal. Secondly, through adversarial training mechanisms to enhance the model's anti-fraud capabilities and establish an edge-cloud collaborative architecture to ensure real-time response, it can effectively identify illegal operations such as water-injected meat, fruits and vegetables with excessive pesticide residues, and tampered labels, thereby improving the safety of people's livelihood products.

[0030] Example 2

[0031] A method for intelligently identifying and managing key commodities in the vegetable basket, used to implement the intelligent identification and management system for key commodities in the vegetable basket of embodiment 1, is characterized by comprising the following steps: S01. Collect multimodal product data, including product images, quality inspection reports, logistics temperature and humidity sensor data, and supplier information, perform data cleaning, standardization, and cloud storage to build a multimodal product database; S02. Based on the multimodal product database, perform pixel-level segmentation and annotation of product images, parse the semantic information of quality inspection reports and associate them with sensor data to generate multi-level quality labels; S03. Train a multimodal fusion deep learning model to simultaneously process image features, quality inspection text, and sensor time series data, and output product category, quality grade, and abnormal risk prediction results; S04. Establish spatiotemporal association rules for multimodal data, dynamically match product image features, quality inspection index fluctuations, and logistics environment changes, and generate a fine-grained product traceability map; S05. Deploy an edge-cloud collaborative monitoring mechanism to perform real-time product image comparison, temperature and humidity deviation warning, and quality degradation prediction. Deploying an edge-cloud collaborative monitoring mechanism includes: S51. Deploy edge computing devices at cold chain transportation nodes and run lightweight detection models to identify package damage in real time. S52. Set multi-level temperature and humidity deviation thresholds: when the temperature deviates by 2°C, an abnormal event is recorded; when it deviates by 5°C, the container door magnetic sensor is activated to verify illegal unpacking behavior; S53. When it is detected that the color characteristics of the product image are inconsistent with the maturity index of the quality inspection report, the batch of products is automatically intercepted from being shipped out and the manual review process is triggered.

[0032] S06. Visualize the traceability information of the entire product chain, the distribution of risk hotspots, and the evidence chain of violations; S07. Iteratively optimize model parameters and traceability rule base logic based on consumer feedback data.

[0033] Specifically, an IoT sensor network deployed at production bases, cold chain vehicles, and point-of-sale terminals captures product images (2-megapixel industrial cameras), environmental data (0.1°C accuracy temperature and humidity sensors), and quality inspection documents in real time. A spatiotemporal alignment algorithm unifies data streams with different sampling frequencies into a common spatiotemporal coordinate system, for example, mapping temperature and humidity curves collected every minute to image frames captured every 5 seconds. The blockchain evidence storage unit employs asymmetric encryption to generate hash values ​​containing timestamps (UTC synchronization) and location information (GPS coordinates) for key data and stores them on the consortium chain. Secondly, an improved Mask R-CNN model incorporates a channel-wise attention mechanism for product image segmentation, achieving a defect detection accuracy of 92.7% mAP. The semantic parsing unit uses the BERT model to extract structured indicators from quality inspection reports and integrates them with sensor data through dynamic weighting. When the transport temperature exceeds a threshold, the weight coefficient of the freshness score for the corresponding batch of products is automatically increased (from 0.3 to 0.6). The three-stream fusion network implements feature interaction through a cross-modal attention layer. For example, when the BiLSTM branch extracts the text feature "insufficient refrigeration time," the TCN branch focuses on analyzing the impact of temperature fluctuations during that period on the degree of browning of fruits and vegetables in the image. Furthermore, the spatiotemporal correlation engine incorporates over 200 expert rules and machine learning-derived rules. When a batch of apples experiences a 5°C temperature rise over two hours during transportation, the system automatically correlates the following features: ① a 15% decrease in peel gloss in the image; ② a 20% deviation from the standard for hardness sensor data; and ③ abnormal ethylene emission in the quality inspection report. Edge computing nodes utilize a quantized MobileNetV3 model (compressed from 85MB to 4.3MB) to achieve real-time detection at 30fps on a Jetson Nano device. Any detection of damaged packaging immediately triggers an audible and visual alarm and uploads a 4K image clip to the cloud for verification. Furthermore, quality feedback from consumers (e.g., "overripe fruit") is converted into model optimization signals through a feature mapping module. If a merchant's mango complaint rate exceeds the threshold for three consecutive days, the system will automatically enhance the detection sensitivity of the fruit stem detachment feature in the batch of images (the IoU threshold is adjusted from 0.5 to 0.7), and at the same time update the association parameters between the origin maturity standard and the transportation time in the traceability rule library.

[0034] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any technician familiar with the present profession can make some changes or modifications to equivalent embodiments of the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An intelligent identification and management system for key commodities in the vegetable basket, characterized by: include: A multimodal data acquisition and preprocessing module, which uses IoT devices to collect product images, temperature and humidity curves, vibration frequencies, and electronic quality inspection reports, and uses blockchain technology to encrypt and store production batches and test values; Intelligent annotation and label generation module, used to perform instance segmentation on product images, parse semantic information in quality inspection reports, and generate multi-level quality labels; The multimodal fusion model training module uses a three-stream fusion network to simultaneously train image, text, and time series data, and outputs product quality prediction results; The spatiotemporal correlation traceability engine module establishes spatiotemporal correlation rules for multimodal data and generates product traceability maps; Edge-cloud collaborative monitoring module deploys edge computing nodes to achieve real-time quality monitoring and multi-level early warning; The visual supervision platform module dynamically displays the full-link traceability information and risk evidence chain of the product; The model iteration optimization module optimizes model parameters and traceability rule base based on consumer feedback data.

2. The intelligent identification and management system for key commodities in the vegetable basket according to claim 1 is characterized in that: The multimodal data acquisition and preprocessing module includes: IoT sensor unit, used to collect images, temperature and humidity curves, vibration frequency data and electronic quality inspection reports of circulation nodes; The spatiotemporal alignment unit performs spatiotemporal alignment on multi-source heterogeneous data and cleans invalid segments; The blockchain evidence storage unit encrypts and stores production batches, transportation tracks, and test values.

3. The intelligent identification and management system for key commodities in the food basket according to claim 1 is characterized in that: The intelligent annotation and label generation module includes: The visual annotation unit uses an improved Mask R-CNN model to perform pixel-level segmentation on product images and mark areas of appearance defects and damaged packaging. Semantic parsing unit, which extracts pesticide residue content and freshness score indicators from quality inspection reports based on natural language processing technology; The label association unit associates image features with quality inspection indicators to generate a three-level quality label system including qualified products, critical products, and non-compliant products.

4. The intelligent identification and management system for key commodities in the food basket according to claim 1 is characterized in that: The multimodal fusion model training module includes: The three-stream fusion network architecture consists of a CNN branch, a BiLSTM branch, and a TCN branch to process images, text, and time series data respectively; Dynamic attention unit, which adjusts feature weights through a cross-modal attention mechanism to identify hidden quality degradation caused by cold chain interruptions; The adversarial training unit generates adversarial samples with water injection, weight increase, and expired label tampering to improve the model's anti-interference ability.

5. The intelligent identification and management system for key commodities in the vegetable basket according to claim 1 is characterized in that: The edge-cloud collaborative monitoring module includes: Edge detection unit, a lightweight model deployed on cold chain vehicles, performs real-time package damage detection and triggers local re-inspection; A multi-level warning unit sets threshold rules for triggering event recording when the temperature deviates by 2°C, and triggering verification of the container door magnetic sensor when the temperature deviates by 5°C; The conflict freezing unit automatically freezes the outbound operation of the batch when the color characteristics of the product image conflict with the quality inspection maturity index.

6. The intelligent identification and management system for key commodities in the food basket according to claim 1 is characterized in that: The spatiotemporal correlation tracing engine module includes: Rule reasoning unit, which establishes spatiotemporal association rules between transport temperature and humidity anomalies and image feature changes; The traceability code generation unit aggregates raw material data from the production end, environmental records from the circulation end, and re-inspection results from the sales end to generate a unique traceability code; The violation mining unit applies graph neural networks to analyze cross-batch data and identify the behavior chain of suppliers replacing origin labels.

7. The intelligent identification and management system for key commodities in the food basket according to claim 1 is characterized in that: The visual supervision platform module includes: A three-dimensional heat map unit shows the spatial correlation between regional complaint rates and transportation temperature and humidity; The intelligent inspection unit automatically associates the supplier's historical violation records with merchants who continuously inspect critical products; The evidence chain encapsulation unit generates tamper-proof regulatory reports based on abnormal image sequences, abnormal sensor data, and quality inspection contradictions.

8. The intelligent identification and management system for key commodities in the vegetable basket according to claim 1 is characterized in that: The model iteration optimization module collects quality evaluation data through the consumer-side feedback interface, uses an incremental learning algorithm to update the multimodal fusion model parameters, and optimizes the traceability rule base logic based on association rule mining.

9. A method for intelligently identifying and managing key commodities in the vegetable basket, for implementing the intelligent identification and management system for key commodities in the vegetable basket according to claims 1-7, characterized in that: The following steps are involved: S01. Collect multimodal product data, including product images, quality inspection reports, logistics temperature and humidity sensor data, and supplier information, perform data cleaning, standardization, and cloud storage to build a multimodal product database; S02. Based on the multimodal product database, perform pixel-level segmentation and annotation on product images, parse semantic information in quality inspection reports and associate them with sensor data to generate multi-level quality labels; S03. Train a multimodal fusion deep learning model to simultaneously process image features, quality inspection text, and sensor time series data, and output product category, quality grade, and abnormal risk prediction results; S04. Establish spatiotemporal association rules for multimodal data, dynamically match product image features, quality inspection index fluctuations, and logistics environment changes, and generate a fine-grained product traceability map; S05. Deploy an edge-cloud collaborative monitoring mechanism to perform real-time product image comparison, temperature and humidity deviation warning, and quality deterioration prediction; S06. Visualize the traceability information of the entire product chain, the distribution of risk hotspots, and the evidence chain of violations; S07. Iteratively optimize model parameters and traceability rule base logic based on consumer feedback data.

10. The method for intelligently identifying and managing key commodities in the food basket according to claim 9, characterized in that: Deploying the edge-cloud collaborative monitoring mechanism in step S05 includes: S51. Deploy edge computing devices at cold chain transportation nodes and run lightweight detection models to identify package damage in real time. S52. Set multi-level temperature and humidity deviation thresholds: when the temperature deviates by 2°C, an abnormal event is recorded; when it deviates by 5°C, the container door magnetic sensor is activated to verify illegal unpacking behavior; S53. When it is detected that the color characteristics of the product image are inconsistent with the maturity index of the quality inspection report, the batch of products is automatically intercepted from leaving the warehouse and the manual review process is triggered.

Images