Agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing

Abstract

The application discloses a non-standard product visual identification and dynamic pricing-based agricultural product matching transaction system, comprising the following modules: an image acquisition module, which establishes a communication connection with a seller client and responds to an image acquisition operation initiated by the seller client; a visual identification processing module, which adopts a pre-trained deep learning model to accurately detect agricultural product quality indexes and convert them into quantifiable feature vectors, combines agricultural product standardization grades, generates a reasonable recommended price interval through a dynamic pricing model, and incorporates agricultural product quality decay rules under logistics time efficiency into the pricing consideration range to protect the reasonable rights and interests of both buyers and sellers, effectively realize the value of high-quality agricultural products, integrate multi-dimensional parameters through a matching transaction module, reduce ineffective communication and intermediate links, improve the transaction rate, retain unalterable whole-process transaction vouchers through a transaction arbitration module to efficiently resolve transaction disputes, constrain the behaviors of buyers and sellers, and maintain a fair and orderly transaction order.

Description

Technical Field

[0001] This application relates to the technical field of agricultural product trading, and in particular to an agricultural product matching trading system based on non-standard product visual recognition and dynamic pricing. Background Technology

[0002] Driven by both consumption upgrades and digital transformation, the agricultural products industry is undergoing a critical shift from traditional extensive development to refined and high-quality development. Agricultural e-commerce is experiencing explosive growth and has become an important engine for promoting the upgrading of the agricultural industry. However, behind the rapid development of the industry, there are still many technical bottlenecks in the trading of non-standard agricultural products.

[0003] Currently, the testing of non-standard agricultural products relies excessively on subjective human judgment or single-dimensional image recognition methods, making it difficult to achieve high-precision comprehensive testing of quality characteristics such as size, color, defects, and maturity. Furthermore, the lack of a dynamic grading standard library that is linked to market demand in real time leads to frequent discrepancies between online transactions and the advertised products, further exacerbating the consumer trust crisis.

[0004] In terms of pricing, non-standard agricultural products still rely on the traditional model of experience-based negotiation or fixed benchmark prices. This fails to capture real-time data on regional market supply and demand fluctuations, making it easy for high-quality products to not command high prices and for prices to deviate from market conditions. Furthermore, the existing pricing system does not consider the quality degradation of agricultural products due to logistics time, and pricing cannot cover the actual costs of the entire transaction cycle. This not only harms the interests of both buyers and sellers but also hinders the realization of the market value of high-quality agricultural products. In addition, the transaction end only adopts a simple information display model, lacking the ability to match supply and demand in a multi-dimensional way. It cannot achieve accurate matching of price, quality, and geographical distance, ultimately leading to problems such as high transaction costs and low transaction success rates.

[0005] Application content

[0006] This application aims to address, at least to some extent, the technical problems in the related art.

[0007] To achieve the above objectives, this application proposes an agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing, comprising the following modules:

[0008] The image acquisition module establishes a communication connection with the seller's client, responds to the image acquisition operation initiated by the seller's client, collects and acquires multi-angle image data of the agricultural product to be traded, and simultaneously collects the basic descriptive information corresponding to the agricultural product.

[0009] The visual recognition processing module is connected to the image acquisition module. Based on a pre-trained deep learning model, it extracts features from the multi-angle image data and generates a quality feature vector of the agricultural product to be traded. The quality feature vector includes appearance size, color uniformity, skin defect rate, and estimated maturity level.

[0010] The non-standard product grading module is connected to the visual recognition processing module. It matches the quality feature vector with a preset non-standard product grading standard library and outputs the standardized grade identifier of the agricultural product to be traded.

[0011] The dynamic pricing engine module acquires real-time market supply and demand data, historical transaction data, and the standardized grade identifier, and uses the dynamic pricing model to calculate the initial recommended price range for the agricultural products to be traded.

[0012] The matching module is connected to the non-standard product grading module and the dynamic pricing engine module, respectively. Based on the standardized grade identifier, the initial recommended price range, and the buyer's client's procurement demand parameters, it generates a matching ranking list and pushes it to the clients of both the buyer and seller.

[0013] The transaction arbitration module stores the original image data acquired by the image acquisition module, the quality feature vector, and the standardized grade identifier as an immutable transaction certificate. In response to a dispute request from either client, it provides complete quality traceability information to the arbitration party or both parties.

[0014] In addition, the application may also include the following additional technical features:

[0015] Specifically, the visual recognition processing module includes: an instance segmentation unit, used to segment the individual contour regions of the agricultural product to be traded from the multi-angle image data; and a three-dimensional reconstruction unit, connected to the instance segmentation submodule, used to reconstruct the three-dimensional model of the agricultural product to be traded based on the multi-view geometric principle to generate the appearance dimensions.

[0016] Specifically, the non-standard product grading module includes: a standard library maintenance unit, which stores and updates a non-standard product grading standard library, containing grading thresholds and judgment rules for the appearance size, color uniformity, skin defect rate, and estimated maturity level of various categories of agricultural products; a feature matching calculation unit, which receives the quality feature vector from the visual recognition processing module, compares each feature parameter with the corresponding threshold in the standard library one by one, and calculates the feature matching degree; and a grade judgment output unit, which determines the standardized grade identifier of agricultural products based on the matching degree and judgment rules, synchronously pushes it to the dynamic pricing engine module and the matching transaction module, and stores it in the transaction arbitration module as evidence.

[0017] Specifically, the dynamic pricing engine module includes: a supply and demand fluctuation analysis unit, used to capture the listing volume, search volume, and transaction rate of agricultural products of the same category within a preset area, and generate a real-time supply and demand coefficient; a loss prediction unit, which calculates the quality decay curve of the agricultural product to be traded under a preset logistics time limit based on the skin defect rate and maturity prediction level in the quality feature vector; and a price generation unit, which is connected to the supply and demand fluctuation analysis unit and the loss prediction unit respectively, and is used to input the supply and demand coefficient and the quality decay curve as correction factors into the dynamic pricing model to adjust the initial recommended price range.

[0018] Specifically, the matching module includes a two-way matching engine, which is used to perform weighted scoring on the standardized level identifier and the initial recommended price range based on the price preference range, quality preference threshold and geographical distance weight in the buyer's procurement demand parameters, and generate the matching sort list arranged in descending order of matching degree.

[0019] Specifically, the transaction arbitration module is built on blockchain technology. It processes the original image data, the quality feature vector, and the standardized grade identifier through hash operation and uploads the hash value to the blockchain network to complete the evidence storage. When a dispute request is triggered, it intuitively displays the entire timeline and key frame comparison diagram of the agricultural product to be traded from image acquisition to grade assessment to the arbitrator, realizing the traceability and visual verification of the transaction process.

[0020] Specifically, it also includes an anti-fraud verification module, which is connected to the image acquisition module and the visual recognition processing module respectively. It is used to detect whether there are deepfake traces, image copying traces or abnormal features that are not currently photographed in the uploaded multi-angle image data, and freeze the transaction permissions of the seller client when an anomaly is detected.

[0021] In summary, the beneficial effects of the agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing proposed in this application are as follows:

[0022] 1. The visual recognition processing module adopts a pre-trained deep learning model to accurately detect agricultural product quality indicators and convert them into quantifiable feature vectors. At the same time, relying on the grading standard library, it flexibly adjusts the grading rules according to market demand and outputs standardized grade labels, so that buyers and sellers can form a unified quality perception, reduce the problem of goods not matching descriptions from the source, and alleviate the crisis of consumer trust.

[0023] 2. By using the dynamic pricing engine module to capture regional market supply and demand and historical transaction data in real time, and combining this with the standardization grade of agricultural products, a reasonable recommended price range is generated through the dynamic pricing model. This incorporates the degradation of agricultural product quality due to logistics time into the pricing consideration, ensuring the reasonable rights and interests of both buyers and sellers and effectively realizing the value of high-quality agricultural products.

[0024] 3. By integrating multi-dimensional parameters through the transaction matching module, accurate two-way matching of core needs such as price, quality, and geographical distance can be achieved, reducing ineffective communication and intermediate links, and improving the transaction rate. At the same time, the transaction arbitration module retains tamper-proof transaction vouchers throughout the entire process, quickly provides quality traceability information, efficiently resolves transaction disputes, constrains the behavior of both buyers and sellers, and maintains a fair and orderly transaction order. Attached Figure Description

[0025] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0026] Figure 1 This is a flowchart of the modules of the agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing in this application;

[0027] Figure 2 This is a flowchart of the image acquisition module and the visual recognition processing module of the agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing in this application.

[0028] Figure 3 This is a flowchart of the dynamic pricing engine module and the matching transaction module of the agricultural product matching transaction system based on non-standard product visual recognition and dynamic pricing in this application. Detailed Implementation

[0029] To make the technical means, inventive features, objectives, and effects of this application easier to understand, the application is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0030] The present application will now be described in further detail with reference to the accompanying drawings.

[0031] As shown in the figure, the agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing in this application embodiment includes the following modules: an image acquisition module, which establishes a communication connection with the seller client, responds to the image acquisition operation initiated by the seller client, acquires and obtains multi-angle image data of the agricultural product to be traded, and synchronously acquires the basic description information corresponding to the agricultural product.

[0032] It should be noted that by establishing a two-way communication link with the seller's client (which can use various communication methods such as wireless LAN and mobile data network), the system responds in real time to the image acquisition commands initiated by the seller. It calls the client's built-in camera or external acquisition device to automatically or manually acquire image data of the agricultural products to be traded according to the preset shooting logic (such as key angles such as front, side, top, and bottom). At the same time, the system guides the seller to enter basic descriptive information of the agricultural products (such as category, origin, planting method, and harvesting time) through the client's interactive interface. The system then associates and binds the image data with the basic descriptive information to generate a unique agricultural product acquisition file, ensuring the traceability of the data.

[0033] This module can capture images from multiple angles to comprehensively and realistically present the appearance of agricultural products, avoiding information omissions caused by shooting from a single angle, and providing comprehensive data support for subsequent visual recognition; it can also collect basic descriptive information simultaneously, realizing the linkage between image data and text information, reducing the workload of sellers to repeatedly enter data, and improving the ease of operation.

[0034] The visual recognition processing module is connected to the image acquisition module. Based on a pre-trained deep learning model, it extracts features from multi-angle image data and generates quality feature vectors for agricultural products to be traded. The quality feature vectors include appearance size, color uniformity, skin defect rate, and maturity prediction level.

[0035] It should be noted that, based on a pre-trained deep learning model (such as a convolutional neural network CNN, specifically optimized for agricultural product image features), the multi-angle image data transmitted by the image acquisition module is preprocessed. Key features in the image are extracted through the model's convolutional layers, pooling layers, and other structures, transforming the image information into quantifiable quality feature vectors. Among these features, the appearance size is calculated using an image segmentation algorithm to determine the pixel ratio of the agricultural product and then converted into actual size parameters. Color uniformity is quantitatively assessed through color histogram analysis. The skin defect rate is identified using an anomaly detection algorithm to determine the defective areas and calculate their proportion. The maturity prediction level is obtained by comparing the image features with a pre-set maturity sample library.

[0036] This module employs a pre-trained deep learning model, which boasts high recognition accuracy and strong anti-interference capabilities. It can effectively distinguish subtle differences in the appearance of agricultural products, avoiding subjective errors caused by human identification. It transforms abstract image information into standardized quality feature vectors, enabling a quantifiable description of agricultural product quality and providing objective and unified data for subsequent grading and dynamic pricing.

[0037] The non-standard product grading module, connected to the visual recognition processing module, matches the quality feature vector with the preset non-standard product grading standard library and outputs the standardized grade label of the agricultural products to be traded.

[0038] Understandably, a non-standard product grading standard library covering different agricultural product categories and quality standards is pre-constructed. This standard library combines industry norms, market demand, historical transaction data, and other factors to set clear grading thresholds and weight allocation rules for the quality feature vectors (appearance size, color uniformity, etc.) of each type of agricultural product. After receiving the quality feature vectors output by the visual recognition processing module, the module compares them with the feature thresholds of each level in the standard library through a similarity matching algorithm, calculates the matching degree based on the weights, and outputs the corresponding standardized grade label (such as level 1, level 2, level 3, etc.).

[0039] This module addresses the pain point of non-standardization in agricultural products. Through standardized grading rules, it transforms the previously difficult-to-define quality of agricultural products into clear grade labels, achieving standardized classification of non-standard products and reducing cognitive biases between buyers and sellers. The grading standard library can be dynamically adjusted according to market changes and category updates, adapting to the quality differences of agricultural products in different regions and seasons, thus improving the system's versatility and flexibility. The automated and objective evaluation process ensures the fairness of transactions while improving the efficiency of grading evaluation, providing rapid support for subsequent transaction matching and pricing.

[0040] The dynamic pricing engine module acquires real-time market supply and demand data, historical transaction data, and standardized grade identifiers, and uses a dynamic pricing model to calculate the initial recommended price range for agricultural products to be traded.

[0041] It should be noted that real-time supply and demand data of the agricultural product market (such as the current supply, demand, and real-time transaction prices of similar agricultural products in the market) and historical transaction data within the system (including transaction prices and volumes of different grades and time periods) are obtained in real time through the data interface. Combined with the standardized grade labels output by the non-standard product grade assessment module, these data are used as input parameters to a preset dynamic pricing model (such as a pricing model based on regression analysis and machine learning, which can automatically adjust model parameters according to market fluctuations). The model calculates a reasonable initial recommended price range for the agricultural product to be traded. The upper and lower limits of the range correspond to the highest and lowest acceptable prices in the market, respectively, taking into account both the seller's profit and the buyer's willingness.

[0042] This module employs a scientifically sound pricing logic, combining real-time supply and demand with historical data to avoid the subjectivity and lag inherent in traditional pricing methods. This ensures that recommended prices better reflect actual market conditions. It uses differentiated pricing based on agricultural product grade designations to achieve premium pricing for high-quality products, fully reflecting their quality value and guaranteeing reasonable profits for sellers. Simultaneously, it provides buyers with transparent and reasonable price references. By outputting price ranges rather than fixed prices, it allows for negotiation between buyers and sellers, enhancing transaction flexibility and success rates. Furthermore, by dynamically adjusting model parameters, it can quickly respond to market fluctuations, ensuring the timeliness and reasonableness of prices.

[0043] The transaction matching module is connected to the non-standard product grading module and the dynamic pricing engine module, respectively. Based on the standardized grade identifier, the initial recommended price range, and the buyer's client's procurement demand parameters, it generates a matching ranking list and pushes it to the clients of both the buyer and seller.

[0044] It should be noted that data interaction is established with the non-standard product grading module and the dynamic pricing engine module respectively. The system obtains the standardized grade identification and initial recommended price range of the agricultural products to be traded in real time. At the same time, it receives the procurement demand parameters submitted by the buyer's client. Through the matching algorithm, the system performs multi-dimensional matching between the seller's agricultural product information and the buyer's procurement demand (prioritizing matching based on core parameters such as grade, price range, and purchase quantity). The matching results are sorted (e.g., by matching degree, price advantage, delivery convenience, etc.) to generate a matching ranking list. The list is then pushed to the corresponding buyer and seller clients through the system push mechanism. The system also supports online communication, negotiation, and transaction confirmation between the two parties.

[0045] This module boasts high matching efficiency, enabling rapid connection between buyers and sellers through automated matching algorithms, reducing intermediaries and improving transaction efficiency. It also offers high matching accuracy, using standardized parameters such as grade identifiers and price ranges to minimize invalid matches and lower communication costs for both parties. Furthermore, it ensures information symmetry by synchronously pushing matching results to both clients, guaranteeing complete transaction information for both sides, enhancing transparency and credibility. Online interaction is also supported, further facilitating transaction completion.

[0046] The transaction arbitration module stores the original image data, quality feature vectors, and standardized grade identifiers acquired by the image acquisition module. As an immutable transaction certificate, it responds to dispute requests from either client and provides complete quality traceability information to the arbitration party or both parties.

[0047] It should be noted that immutable data storage technologies (such as blockchain and distributed ledgers) are used to permanently store key data such as the original image data acquired by the image acquisition module, the quality feature vector generated by the visual recognition processing module, and the standardized grade label output by the non-standard product grade assessment module, forming a complete chain of transaction vouchers to ensure that the data cannot be tampered with or forged. When either client initiates a transaction dispute (such as the buyer believing that the received agricultural products do not match the grade label, or the seller believing that the buyer has maliciously breached the contract), the module responds to the dispute request, extracts the corresponding complete quality traceability information (including the original image, feature vector, grade assessment basis, etc.) from the storage system, and pushes it simultaneously to the arbitration party (such as the system platform arbitrator) and both parties to the transaction, providing objective and effective evidence support for dispute resolution.

[0048] This module provides a traceable and tamper-proof chain of evidence for transaction disputes, reducing the cost of handling such disputes. The stored original data and traceability information can fully reconstruct the quality assessment and transaction process of agricultural products, protecting the legitimate rights and interests of both buyers and sellers and enhancing their trust in the trading platform. No manual evidence collection or submission is required, dispute responses are rapid, the arbitration process can be efficiently advanced, the impact of transaction disputes on transaction progress can be reduced, and the stability of the transaction order can be maintained.

[0049] In one embodiment of this application, the visual recognition processing module includes: an instance segmentation unit for segmenting the individual outline region of the agricultural product to be traded from multi-angle image data; and a three-dimensional reconstruction unit connected to the instance segmentation submodule for reconstructing a three-dimensional model of the agricultural product to be traded based on multi-view geometric principles to generate the appearance dimensions.

[0050] Specifically, the instance segmentation unit is used to receive multi-angle image data transmitted by the image acquisition module. Through a deep learning-based instance segmentation algorithm, it accurately segments the individual contour regions of the agricultural products to be traded, effectively eliminating background clutter, other agricultural products, and interference from the shooting environment, while preserving the edge details of the agricultural products, laying the foundation for subsequent size measurement and defect identification.

[0051] The 3D reconstruction unit and the instance segmentation unit establish a real-time data connection. Based on the multi-view geometry principle, feature points of the agricultural product outline under each view are extracted and matched and triangulated to reconstruct a high-precision 3D model of the agricultural product to be traded. The actual appearance size of the agricultural product is generated by converting the model coordinates, realizing non-contact and accurate measurement of the agricultural product size, avoiding errors and losses caused by manual measurement.

[0052] In one embodiment of this application, the non-standard product grading module includes: a standard library maintenance unit, which stores and updates a non-standard product grading standard library, which contains grading thresholds and judgment rules for the appearance size, color uniformity, skin defect rate, and maturity prediction grades of various categories of agricultural products; a feature matching calculation unit, which receives the quality feature vector from the visual recognition processing module, compares each feature parameter with the corresponding threshold in the standard library one by one, and calculates the feature matching degree; and a grade judgment output unit, which determines the standardized grade identifier of agricultural products according to the matching degree and judgment rules, pushes it synchronously to the dynamic pricing engine module and the matching transaction module, and stores it in the transaction arbitration module as evidence.

[0053] It should be noted that the standard library maintenance unit is responsible for storing and dynamically updating the non-standard product grading standard library. The library has preset grading thresholds and judgment rules for key indicators such as appearance size, color uniformity, skin defect rate (including defect area, defect type, defect location), and maturity level for different categories of agricultural products (such as vegetables, fruits, and livestock products). The grading thresholds can be flexibly adjusted according to market demand, industry standards, and the characteristics of agricultural product origin. The judgment rules adopt a multi-feature weighted comprehensive judgment method to avoid misjudgment based on a single indicator.

[0054] The feature matching operation unit receives the agricultural product quality feature vectors output by the visual recognition processing module in real time, compares each feature parameter with the corresponding category grading threshold in the standard library one by one, and generates a quantitative feature matching score by calculating parameter deviation value, feature fit, etc.

[0055] The grade determination output unit determines the standardized grade identifier (such as Grade 1, Grade 2, Grade 3, or substandard product) of the agricultural product to be traded based on the feature matching score and preset determination rules. At the same time, it pushes the grade identifier to the dynamic pricing engine module and the matching transaction module through the system interface to provide a basis for pricing and matching. It also stores the grade determination results and related feature data in the transaction arbitration module as evidence for any potential transaction disputes and ensures the traceability of the grade assessment.

[0056] In one embodiment of this application, the dynamic pricing engine module includes: a supply and demand fluctuation analysis unit, used to capture the listing volume, search volume and transaction rate of agricultural products of the same category within a preset area, and generate a real-time supply and demand coefficient; a loss prediction unit, used to calculate the quality decay curve of the agricultural product to be traded under a preset logistics time limit based on the skin defect rate and maturity prediction level in the quality feature vector; and a price generation unit, connected to the supply and demand fluctuation analysis unit and the loss prediction unit respectively, used to input the supply and demand coefficient and the quality decay curve as correction factors into the dynamic pricing model to adjust the initial recommended price range.

[0057] Understandably, the supply and demand fluctuation analysis unit captures real-time data such as the number of agricultural products listed, user search volume, historical transaction rate, and inventory turnover rate of the same category within a preset area through the system interface. Through data normalization processing and statistical analysis, it generates a real-time supply and demand coefficient. A coefficient greater than 1 indicates that demand exceeds supply, and a coefficient less than 1 indicates that supply exceeds demand, which intuitively reflects the market supply and demand fluctuation situation.

[0058] The loss prediction unit, based on the skin defect rate and maturity estimate level in the quality feature vector output by the visual recognition processing module, combined with the preset logistics timeliness and referring to historical loss data, calculates the quality decay curve of the agricultural products to be traded during transportation and storage by fitting a loss model, and predicts the quality changes and loss rate at different time points.

[0059] The price generation unit establishes data connections with the supply and demand fluctuation analysis unit and the loss prediction unit, respectively. It uses the supply and demand coefficient and the quality decay curve as core correction factors and inputs them into a preset dynamic pricing model (constructed by combining cost price and market benchmark price). The model calculates and adjusts the initial recommended price range to achieve a higher price for better quality and tighter supply and demand, and a lower price for higher loss rate and looser supply and demand, thus balancing the interests of both buyers and sellers.

[0060] In one embodiment of this application, the matching module includes a two-way matching engine. The two-way matching engine is used to perform weighted scoring on the standardized level identifier and the initial recommended price range based on the price preference range, quality preference threshold and geographical distance weight in the buyer's procurement demand parameters, and generate a matching sorting list arranged in descending order of matching degree.

[0061] It should be noted that the two-way matching engine first receives the procurement demand parameters submitted by the buyer's client, including price preference range (such as the highest acceptable price and expected price range), quality preference threshold (such as the minimum grade requirement and the upper limit of acceptable defect rate), and geographical distance weight (such as prioritizing suppliers that are closer to reduce logistics costs and losses). Then, it extracts the standardized grade identifier of the agricultural products to be traded and the initial recommended price range output by the dynamic pricing engine module.

[0062] Subsequently, the procurement demand parameters and agricultural product information are weighted and scored from multiple dimensions, with corresponding weights assigned to price matching degree, quality matching degree, and geographical distance matching degree (which can be adjusted according to system settings or user needs). Finally, a matching sorting list is generated in descending order of matching degree and pushed to the buyer's client for selection. At the same time, the matching results are synchronously fed back to the corresponding seller's client to achieve two-way accurate matching.

[0063] In one embodiment of this application, the transaction arbitration module is built on blockchain technology. It processes the original image data, quality feature vector and standardized grade identifier through hash operation and uploads the hash value to the blockchain network to complete the evidence storage. When a dispute request is triggered, it intuitively displays the timeline and key frame comparison diagram of the entire process of the agricultural product to be traded from image acquisition to grade assessment to the arbitrator, so as to realize the traceability and visual verification of the transaction process.

[0064] Specifically, when a transaction dispute is triggered by both the buyer and seller (e.g., the buyer believes that the received agricultural products do not match the grade label), the transaction arbitration module can retrieve the evidence data from the blockchain and intuitively display the entire timeline of the agricultural products to be traded, from image acquisition, instance segmentation, 3D reconstruction, grade assessment to pricing and matching, to the arbitrator. At the same time, it can display key frame comparison diagrams of each key link, realizing full-link traceability and visual verification of the transaction process.

[0065] In one embodiment of this application, an anti-fraud verification module is also included. The anti-fraud verification module is connected to the image acquisition module and the visual recognition processing module respectively. It is used to detect whether there are deepfake traces, image copy traces or abnormal features that are not currently photographed in the uploaded multi-angle image data, and freeze the transaction permissions of the seller client when an anomaly is detected.

[0066] It should be noted that the anti-fraud verification module performs real-time fraud detection on multi-angle image data uploaded by the seller's client. The specific detection content includes whether there are traces of deepfake (such as fake agricultural product images generated by AI), traces of image duplication (such as repeated use of images of other agricultural products), or abnormal features that are not currently taken of the actual object (such as the image not matching the current time and location, or the image showing signs of splicing or retouching).

[0067] A deep learning-based fraud detection model is used to make a comprehensive judgment by extracting texture features, lighting features, timestamp information, etc. of images. When abnormal features are detected, a security warning is immediately triggered, the seller's client transaction permissions are frozen, and the seller is prohibited from publishing new transaction information and participating in transaction matching. At the same time, an abnormal notification is sent to the platform management. The transaction permissions can only be restored after the management personnel verify that there are no errors and the seller completes rectification. This effectively prevents fraudulent activities such as fake agricultural product transactions and selling inferior products as superior ones.

[0068] It should be noted that, in this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0069] The present application and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present application. The actual structure is not limited to this. In conclusion, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present application, such design should fall within the protection scope of the present application.

Claims

1. A non-standard visual identification and dynamic pricing-based agricultural product matchmaking transaction system, characterized in that, Includes the following modules: The image acquisition module establishes a communication connection with the seller's client, responds to the image acquisition operation initiated by the seller's client, acquires and obtains multi-angle image data of the agricultural product to be traded, and simultaneously acquires the basic descriptive information corresponding to the agricultural product. The visual recognition processing module is connected to the image acquisition module. Based on a pre-trained deep learning model, it extracts features from the multi-angle image data and generates a quality feature vector of the agricultural product to be traded. The quality feature vector includes appearance size, color uniformity, skin defect rate and maturity prediction level. The non-standard product grading module is connected to the visual recognition processing module. It matches the quality feature vector with a preset non-standard product grading standard library and outputs the standardized grade identifier of the agricultural product to be traded. The dynamic pricing engine module acquires real-time market supply and demand data, historical transaction data, and the standardized grade identifier, and uses the dynamic pricing model to calculate the initial recommended price range for the agricultural products to be traded. The matching transaction module is connected to the non-standard product grade assessment module and the dynamic pricing engine module respectively. Based on the standardization grade identifier, the initial recommended price range and the buyer's client's procurement demand parameters, it generates a matching ranking list and pushes it to the buyer and seller's clients. The transaction arbitration module stores the original image data acquired by the image acquisition module, the quality feature vector, and the standardized grade identifier as an immutable transaction certificate. In response to a dispute request from either client, it provides complete quality traceability information to the arbitration party or both parties.

2. The non-commodity based visual identification and dynamic pricing facilitated trading system for agricultural produce as claimed in claim 1 wherein, The visual recognition processing module includes: An instance segmentation unit is used to segment the individual contour regions of the agricultural product to be traded from the multi-angle image data; The three-dimensional reconstruction unit, connected to the instance segmentation submodule, reconstructs the three-dimensional model of the agricultural product to be traded based on the multi-view geometric principle to generate the appearance dimensions.

3. The non-commodity based visual identification and dynamic pricing facilitated trading system for agricultural produce as claimed in claim 1 wherein, The non-standard product grade assessment module includes: The standard library maintenance unit stores and updates the non-standard product grading standard library, which contains grading thresholds and judgment rules for the appearance size, color uniformity, skin defect rate, and estimated maturity level of various agricultural products. The feature matching operation unit receives the quality feature vector from the visual recognition processing module, compares each feature parameter with the corresponding threshold in the standard library one by one, and calculates the feature matching degree. The grade determination output unit determines the standardization grade label of agricultural products based on the matching degree and determination rules, and pushes it to the dynamic pricing engine module and the matching transaction module simultaneously, and stores it in the transaction arbitration module as evidence.

4. The non-commodity based visual identification and dynamic pricing facilitated trading system for agricultural produce as claimed in claim 1 wherein, The dynamic pricing engine module includes: The supply and demand fluctuation analysis unit is used to capture the listing volume, search volume and transaction rate of the same type of agricultural products in a preset area and generate real-time supply and demand coefficients. The loss prediction unit calculates the quality decay curve of the agricultural product to be traded under a preset logistics time limit based on the skin defect rate and maturity prediction level in the quality feature vector. The price generation unit is connected to the supply and demand fluctuation analysis unit and the loss prediction unit, respectively, and is used to input the supply and demand coefficient and the quality decay curve as correction factors into the dynamic pricing model to adjust the initial recommended price range.

5. The agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing according to claim 1, characterized in that, The matching module includes a two-way matching engine, which is used to perform weighted scoring on the standardized level identifier and the initial recommended price range based on the price preference range, quality preference threshold and geographical distance weight in the buyer's procurement demand parameters, and generate the matching sort list arranged in descending order of matching degree.

6. The agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing according to claim 1, characterized in that, The transaction arbitration module is built on blockchain technology. It processes the original image data, the quality feature vector, and the standardized grade identifier through hash operation and uploads the hash value to the blockchain network to complete the evidence storage. When a dispute request is triggered, it intuitively displays the entire timeline and key frame comparison diagram of the agricultural product to be traded from image acquisition to grade assessment to the arbitrator, realizing the traceability and visual verification of the transaction process.

7. The agricultural product matching and trading system based on non-standard product visual recognition and dynamic pricing according to claim 1, characterized in that, It also includes an anti-fraud verification module, which is connected to the image acquisition module and the visual recognition processing module respectively. It is used to detect whether there are traces of deepfake, traces of image duplication or abnormal features that are not currently photographed in the uploaded multi-angle image data, and freeze the transaction permissions of the seller client when an anomaly is detected.

Images