Goods pledge management method based on AI vision and related device

Abstract

The embodiment of the invention relates to the field of data processing, and provides a pledged cargo management method based on AI vision and a related device, and the method comprises the steps: carrying out the multi-dimensional vision monitoring of a current pledged cargo based on an AI vision technology, and obtaining the multi-dimensional vision monitoring data; according to the market value multi-source data and the multi-dimensional visual monitoring data, constructing a target market-goods dual-channel valuation model; performing market value estimation on the current pledged goods based on a target market-goods dual-channel value estimation model to obtain current market value data; when the current market value data meets a preset condition, generating pledged cargo risk early warning information; and carrying out risk early warning according to the pledged cargo risk early warning information. By adopting the method provided by the invention, more accurate and more effective risk management of the current pledged goods can be realized based on the AI vision technology.

Description

Technical Field

[0001] This application relates to the technical fields of data processing, fintech, and warehousing supervision, and particularly relates to a method and related device for managing pledged goods based on AI vision. Background Art

[0002] Currently, in the current financial and warehousing business fields, the status quo of pledged goods management is not optimistic, and there are many problems of improper management, which have caused varying degrees of negative impacts on financial institutions, cargo owners, and warehousing enterprises. In warehouse management, manual inventory of goods is a common method. However, this method is extremely inefficient. Especially in large warehouses, when faced with a wide variety and large quantity of mortgaged goods, it takes a long time to manually count them one by one. Moreover, manual operations are easily interfered by subjective factors such as fatigue and negligence, resulting in inaccurate inventory results, and the phenomenon of the quantity of goods not matching the actual situation often occurs. In addition, the warehouse supervision system also has defects. The layout of monitoring equipment is unreasonable, there are many monitoring blind spots, and the status of goods cannot be comprehensively and real-time grasped. The security measures of some warehouses are insufficient, and it is difficult to discover and trace the responsibility in a timely manner after the goods are stolen or damaged, bringing direct economic losses to cargo owners and financial institutions. Therefore, how to achieve more effective risk management of pledged goods has become an urgent problem to be solved currently. Summary of the Invention

[0003] The embodiments of this application provide a method and related device for managing pledged goods based on AI vision, which can realize more accurate and effective risk management of current pledged goods based on AI vision technology.

[0004] The first aspect of the embodiments of this application provides a method for managing pledged goods based on AI vision, and the method includes: Based on AI vision technology, perform multi-dimensional visual monitoring on current pledged goods to obtain multi-dimensional visual monitoring data; Construct a target market-goods dual-channel valuation model according to multi-source data of market value and multi-dimensional visual monitoring data; Based on the target market-goods dual-channel valuation model, estimate the market value of current pledged goods to obtain current market value data; When the current market value data meets the preset conditions, generate a risk warning information for pledged goods; Perform risk warning according to the risk warning information for pledged goods.

[0005] In this example, based on AI vision technology, multi-dimensional visual monitoring is performed on the current pledged goods to obtain multi-dimensional visual monitoring data. Then, based on the multi-source data of market value and the multi-dimensional visual monitoring data, a target market-goods dual-channel valuation model is constructed. Thus, based on the target market-goods dual-channel valuation model, the market value of the current pledged goods can be estimated to obtain the current market value data. Further, when the current market value data meets the preset conditions, a risk warning message for the pledged goods is generated. Furthermore, based on the risk warning message for the pledged goods, risk warning can be carried out, enabling more accurate and effective risk management of the current pledged goods.

[0006] In a second aspect of the embodiments of the present application, a pledged goods management device based on AI vision is provided. The device includes: A first processing unit, configured to perform multi-dimensional visual monitoring on the current pledged goods based on AI vision technology to obtain multi-dimensional visual monitoring data; A second processing unit, configured to construct a target market-goods dual-channel valuation model according to the multi-source data of market value and the multi-dimensional visual monitoring data; A third processing unit, configured to estimate the market value of the current pledged goods based on the target market-goods dual-channel valuation model to obtain the current market value data; A fourth processing unit, configured to generate a risk warning message for the pledged goods when the current market value data meets the preset conditions; A fifth processing unit, configured to perform risk warning based on the risk warning message for the pledged goods.

[0007] In a third aspect of the embodiments of the present application, a terminal is provided, including a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. Among them, the memory is used to store a computer program, and the computer program includes program instructions. The processor is configured to call the program instructions to execute the step instructions as described in the first aspect of the embodiments of the present application.

[0008] In a fourth aspect of the embodiments of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program for electronic data exchange. The computer program enables a computer to execute some or all of the steps described in the first aspect of the embodiments of the present application.

[0009] A fifth aspect of the embodiments of the present application provides a computer program product, where the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute some or all of the steps described in the first aspect of the embodiments of the present application. The computer program product may be a software installation package. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0011] Figure 1 FIG. is a schematic flow chart of a method for managing pledged goods based on AI vision provided by an embodiment of the present application; Figure 2 FIG. is a schematic structural diagram of a terminal provided by an embodiment of the present application; Figure 3 FIG. is a schematic structural diagram of a device for managing pledged goods based on AI vision provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0012] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present application belong to the scope of protection of the present application.

[0013] The terms "first", "second", etc. in the specification and claims of the present application and the above drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0014] References to "embodiments" in this application mean that specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0015] To better understand a method for managing pledged goods based on AI vision provided in an embodiment of this application, the following first briefly introduces the defects of existing pledged goods management. Currently, there are the following problems in existing pledged goods management: 1. Dependence on manual inspections: For example, relying on on-site personnel to regularly verify the quantity and status of goods is not only inefficient but also prone to missing abnormalities (such as damaged or swapped goods); 2. Lagging value assessment: For example, the valuation of pledged goods is mostly based on static warehousing data and it is difficult to reflect market price fluctuations or physical losses of goods in a timely manner; 3. Insufficient risk warning: For example, traditional systems only provide basic inventory data and lack the ability to intelligently analyze the dynamic deviation of the value of pledged goods; 4. Low data credibility: The data of the goods owner and financial institutions are not synchronized, which is prone to information asymmetry risks.

[0016] The embodiment of this application aims to solve the above problems and provides a method for managing pledged goods based on AI vision. This method can perform multi-dimensional visual monitoring on current pledged goods through artificial intelligence (AI) vision technology, and by constructing a target market-goods dual-channel valuation model, to understand the current market value of pledged goods in real time. Thus, when the fluctuation of this current market value is greater than a preset condition, a risk warning information for pledged goods is generated and risk warning is carried out, and the purpose of more accurate and effective risk management for current pledged goods can be achieved.

[0017] Please refer to Figure 1 , Figure 1 which is a schematic flowchart of a method for managing pledged goods based on AI vision provided in an embodiment of this application. This method includes: S10: Based on AI vision technology, perform multi-dimensional visual monitoring on current pledged goods to obtain multi-dimensional visual monitoring data.

[0018] Among them, AI vision technology can be understood as using technical means such as computer vision algorithms and deep learning models to enable a computer to understand and analyze the content in an image or video, to simulate human visual perception and understanding capabilities, in order to achieve tasks such as object recognition, detection, tracking, and analysis.

[0019] Multi-dimensional visual monitoring can be understood as monitoring the pledged goods from multiple perspectives and in multiple ways. Specifically, it can include, but is not limited to, using different types of visual devices, such as panoramic cameras to obtain the overall appearance and quantity information of the goods, and three-dimensional (3D) laser scanners to obtain the three-dimensional spatial data of the goods. Optionally, the monitoring content can include multiple dimensions such as the quantity, volume, shape, and surface condition of the goods, and this application does not limit this.

[0020] Multi-dimensional visual monitoring data can be understood as various data about the pledged goods collected through multi-dimensional visual monitoring. The multi-dimensional visual monitoring data can include, but is not limited to, information such as the real-time quantity, volume size, shape change, and whether there is damage on the surface of the goods, and can exist in forms such as images, point cloud data, and numerical values, and this application does not limit this.

[0021] It should be understood that based on AI vision technology, performing multi-dimensional visual monitoring on the current pledged goods to obtain multi-dimensional visual monitoring data refers to the process of how to use AI vision technology to monitor the current pledged goods in different dimensions to obtain multi-dimensional visual monitoring data. Among them, in step S10, that is, based on AI vision technology, performing multi-dimensional visual monitoring on the current pledged goods to obtain multi-dimensional visual monitoring data, can include the following steps: S11: Based on AI vision technology, collect the current pledged state image of the current pledged goods; S12: Perform identification processing on the current pledged state image for the pledged goods to obtain dynamic quantity statistical data of the pledged goods; S13: Based on AI vision technology, obtain the current pledged point cloud data of the current pledged goods; S14: Use point cloud reconstruction technology to perform three-dimensional model construction processing on the current pledged point cloud data to obtain volume and shape monitoring data of the pledged goods; S15: Construct a volume-weight correlation model based on the volume and shape monitoring data, and obtain reference dynamic weight estimation data through the volume-weight correlation model; S16: Extract density-related feature data based on AI vision technology, and perform non-linear correction on the reference dynamic weight estimation data based on the density-related feature data to obtain dynamic weight estimation data of the pledged goods; S17: Determine multi-dimensional visual monitoring data according to the dynamic quantity statistical data of the pledged goods, the volume and shape monitoring data of the pledged goods, and the dynamic weight estimation data of the pledged goods.

[0022] Among them, devices such as panoramic cameras and depth cameras deployed in the warehouse can be used to collect images of the current pledged state of the current pledged goods to obtain the current pledged state image. That is to say, the appearance image of the pledged goods can be captured in real time. By obtaining the initial visual features of the goods, such as stacking form, packaging integrity, surface defects, etc., it can provide raw data for subsequent analysis. For example, during the pledge of steel, the degree of rust of the steel can be identified through image recognition; during the pledge of agricultural products, it can be detected whether the packaging is damaged or damp through the above methods.

[0023] Object detection algorithms can be adopted to perform identification processing on the current pledged state image of the pledged goods to obtain dynamic quantity statistical data of the pledged goods. The object detection algorithm can be computer vision technology based on deep learning, such as You Only Look Once (YOLO), Faster Region-based Convolutional Neural Network (Faster R-CNN). The object recognition model can be obtained through training to identify relevant information such as the category and location of the pledged goods based on the object recognition model.

[0024] The dynamic quantity statistical data of the pledged goods can be understood as the dynamic change data used to reflect the quantity of the pledged goods obtained by comparing the current pledged state images of consecutive frames and monitoring the change in the quantity of the goods in real time. For example, during the pledge of oil products, by identifying the pledged state images of oil barrels in consecutive frames, the change in the number of oil barrels can be obtained, and then it can be determined whether there is abnormal out-of-storage in subsequent steps. This application does not limit this.

[0025] Devices such as 3D laser scanners or structured light cameras deployed in the warehouse can be used to obtain a set of three-dimensional coordinate points on the surface of the current pledged goods to obtain the current pledged point cloud data that can reflect the spatial set information of the object. Further, through point cloud reconstruction technology, the discrete point cloud data can be reconstructed into a complete three-dimensional model through algorithms such as interpolation and meshing. For example, during the monitoring of coal piles, through point cloud reconstruction technology, a complete three-dimensional model of the coal pile can be constructed to calculate the volume change of the coal pile. By constructing a three-dimensional model, the changes in the volume and shape of the pledged goods can be intuitively displayed, so that the multi-faceted state of the current pledged goods can be more comprehensively understood.

[0026] A linear regression algorithm can be adopted to construct a volume-weight correlation model based on the volume and morphology monitoring data. Then, the weight of the current pledged goods can be estimated through this volume-weight correlation model to obtain reference dynamic weight estimation data. Further, the reference dynamic weight estimation data can be non-linearly corrected by combining density-related feature data (such as stacking density, porosity, etc.) identified by AI vision to further improve the accuracy of the weight estimation for the pledged goods. Specifically, non-linear models such as neural networks can be used for correction. For example, the density-related feature data can be used as input to adjust the reference dynamic weight estimation data. Through the above non-linear correction, the complex changes in density in the actual situation can be better adapted, and the accuracy of weight estimation can be improved.

[0027] Further, multi-source data such as the dynamic quantity statistics data of the pledged goods, the volume and morphology monitoring data of the pledged goods, and the dynamic weight estimation data of the pledged goods are aligned in space-time and feature-fused. For example, by using a weighted fusion method, etc., a comprehensive portrait of the goods status can be formed, and structured data reports or visual image data forms can be output to better support real-time monitoring and risk warning in the subsequent steps.

[0028] Exemplarily, taking the scenario of steel material pledge supervision as an example, through image acquisition and processing, such as obtaining real-time images of the steel stack through a high-definition camera; and through quantity statistics processing, such as using an object detection algorithm to identify the number of steel bars of different specifications and comparing them with the warehousing records; three-dimensional modeling processing can be carried out, such as using point cloud data to reconstruct the three-dimensional model of the steel stack to further calculate the volume change rate; weight estimation processing can be carried out, such as estimating the total weight based on the steel density and volume and correcting the density parameter through the surface rust degree; and then risk warning can be carried out in the subsequent steps. For example, when the quantity, volume, or weight changes exceed the preset threshold, an abnormal alarm can be triggered. This application does not limit this.

[0029] S20: Construct a target market-goods dual-channel valuation model based on the multi-source market value data and the multi-dimensional visual monitoring data.

[0030] The multi-source market value data can be understood as data related to the market value of the pledged goods from multiple channels and various types. This multi-source market value data can include but is not limited to real-time transaction data on various commodity and industry price index platforms to reflect market price fluctuations; key event data in news and industry reports to be used for analyzing market supply and demand and trend changes; and macroeconomic data, policy and regulation information, etc. to reflect the impact on the market value of the pledged goods from different levels.

[0031] The target market - dual - channel valuation model for goods can be understood as a valuation model that comprehensively considers market factors and the physical state of the goods. The "dual - channel" can refer to the analysis and calculation from two dimensions: the market conditions and the physical state of the goods. By combining multi - source data of market value to reflect market - level impacts such as market price fluctuations and supply - demand relationships, and multi - dimensional visual monitoring data to reflect changes in the physical state of the goods, such as quantity, volume, and shape, a model that can more accurately evaluate the current value of the pledged goods can be constructed, namely the target market - dual - channel valuation model for goods.

[0032] It should be understood that constructing the target market - dual - channel valuation model based on the multi - source data of market value and the multi - dimensional visual monitoring data refers to the process of how to construct the target market - dual - channel valuation model. Among them, in step S20, that is, constructing the target market - dual - channel valuation model based on the multi - source data of market value and the multi - dimensional visual monitoring data, the following steps can be included: S21: Pre - process the multi - source data of market value to obtain target multi - source data of market value; S22: Evaluate the impact of supply - demand public opinion based on the target multi - source data of market value to obtain quantified data on the impact of supply - demand public opinion; S23: Based on the initial model, obtain a reference market - dual - channel valuation model according to the quantified data on the impact of supply - demand public opinion and the multi - dimensional visual monitoring data; S24: Train and optimize the reference market - dual - channel valuation model to obtain the target market - dual - channel valuation model.

[0033] Since the multi - source data of market value contains data related to the market value of the pledged goods collected from various channels, problems such as data missing, errors, and inconsistent formats may exist. Pre - processing can clean, transform, and integrate the multi - source data of market value to improve data quality and make it more suitable for subsequent analysis. Through data cleaning, duplicate data is removed, incorrect data is corrected, and missing values are filled; data in different formats are uniformly transformed, such as standardizing the price data formats of different platforms; multi - source data is integrated, and data from different channels but related to the same pledged goods is associated. Finally, the target multi - source data of market value can be obtained, which can better provide a reliable data basis for subsequent analysis.

[0034] Based on the multi-source data of the target market value after preprocessing, analyze the degree of influence of the public opinion information related to the supply and demand of the pledged goods on its market value, and convert this influence into quantifiable data, so as to obtain the quantifiable data of the influence of supply and demand public opinion. Specifically, the public opinion information related to supply and demand can be extracted from the multi-source data of the target market value first, such as events like industry production restriction policies, logistics interruptions, and the emergence of new market demands; then use the sentiment analysis algorithm in natural language processing technology to judge the sentiment tendency (whether it is favorable or unfavorable) of these public opinion events on the market value of the pledged goods; then combine technologies such as keyword extraction and text clustering to determine the importance and influence of each public opinion event, and finally quantify these influences to obtain the quantifiable data of the influence of supply and demand public opinion, which can more accurately reflect the influence of market public opinion on the value of the pledged goods.

[0035] It should be understood that evaluating the influence of supply and demand public opinion based on the multi-source data of the target market value and obtaining the quantifiable data of the influence of supply and demand public opinion refers to the process of how to evaluate the influence of supply and demand public opinion and obtain the quantifiable data of the influence of supply and demand public opinion. Among them, in step S22, that is, evaluating the influence of supply and demand public opinion based on the multi-source data of the target market value and obtaining the quantifiable data of the influence of supply and demand public opinion, may include the following steps: S221: Based on the multi-source data of the target market value, establish and obtain a keyword library for pledged goods for different types of pledged goods; S222: Judge the sentiment tendency of public opinion events in the multi-source data of the target market value to obtain sentiment tendency influence data; S223: Analyze the influence of each public opinion event according to the keyword library of pledged goods to obtain the influence weight of public opinion events; S224: Based on the sentiment tendency influence data, the influence weight of public opinion events, and the spread range data of each public opinion event, quantify the value influence of each public opinion event on the pledged goods to obtain the quantifiable data of the influence of supply and demand public opinion.

[0036] Professional terms, industry vocabulary, and event labels related to the pledged goods can be extracted from the multi-source data of the target market value to form a standardized thesaurus, so as to obtain the keyword library of pledged goods. Specifically, domain word segmentation processing can be adopted, such as performing domain-specific word segmentation for the types of pledged goods (such as steel, grain, chemical products, etc.); entity recognition processing can be adopted, such as identifying key entities such as enterprise names, policy documents, and port names through Named Entity Recognition (NER) technology; word frequency statistics processing can be adopted, such as screening high-frequency and economically significant words (such as production restriction orders, inventory backlogs, trade wars, etc.). By establishing a keyword library for pledged goods, semantic anchors can be provided for subsequent public opinion analysis and the accuracy of subsequent sentiment analysis and influence evaluation can be ensured.

[0037] Natural language processing (NLP) techniques can be used to perform sentiment polarity analysis on public opinion texts to determine the direction of their impact on the value of the pledged assets (such as positive / negative). Specifically, the sentiment dictionary method can be adopted, such as constructing a sentiment dictionary in the financial field (for example, "rise" and "shortage" can be positive words, and "plunge" and "surplus" can be negative words); machine learning models can be used, such as using pre-trained models like BERT for sentiment classification to output sentiment scores (such as in the range of -1 to +1); multimodal fusion can be adopted, such as combining text sentiment and numerical indicators (such as price volatility) to enhance the judgment accuracy; and then the sentiment tendency value corresponding to each public opinion event can be output (for example, +0.8 indicates strong positive, and -0.6 indicates negative). This application does not limit this.

[0038] In a possible implementation, the degree of impact of each public opinion event on the value of the pledged assets can be further evaluated, such as considering the event type, timeliness, and relevance. Specifically, the relevance of each public opinion event can be evaluated according to the keyword library of the pledged goods to obtain the keyword relevance; the weight of the event type of each public opinion event can be calculated according to the keyword library of the pledged goods to obtain the weight of the public opinion event type; the timeliness of each public opinion event can be evaluated to obtain the timeliness factor; and the impact weight of the public opinion event can be determined according to the keyword relevance, the weight of the public opinion event type, and the timeliness factor.

[0039] Specifically, the weight of the public opinion event type can be calculated, such as determining that different types of public opinion events have different degrees of impact on the value of the pledged assets (such as policy changes > industry conferences > corporate news); the timeliness factor can be calculated, such as using a time window function to assign higher weights to recent events (for example, the weight of events within 3 days is 1, and the weight of events outside 7 days is 0.3); the keyword relevance can be calculated, such as calculating the semantic similarity between the event and the pledged assets based on the keyword library. For example, the weight of events related to "rebars" can be higher than that of "non-ferrous metals". Optionally, the final impact weight of the public opinion event can be the product of the time type weight, the timeliness factor, and the keyword relevance, that is, the impact weight of the public opinion event = the weight of the public opinion event type Timeliness factor Keyword relevance. This application does not limit this.

[0040] Optionally, the timeliness factor can be understood as an exponential decay function of the event occurrence time and the current time. The specific formula can be seen as follows: where is the timeliness factor (i.e., timeliness intensity) of the public opinion event at time t, and the value range is (0, 1]; is the time interval between the occurrence time of the public opinion event and the current time (usually measured in time units such as days, hours, minutes, etc.), ; is the natural constant; is the attenuation coefficient (or the "attenuation rate parameter" of the timeliness factor), is a positive number.

[0041] It can be understood that as time goes by, the impact or importance of the public opinion event will gradually attenuate, the closer it is to 0, the weaker the timeliness, the closer it is to 1, the stronger the timeliness. For example, when, it means that the timeliness is the strongest when the event just occurs.

[0042] By combining the sentiment tendency impact data, the public opinion event impact weight, and the spread range data of each public opinion event, the supply-demand public opinion impact quantification data can be formed. This supply-demand public opinion impact quantification data is a quantification index that can be directly used in the target market - goods dual-channel valuation model. The spread range data can be data indicating the spread degree of the public opinion event obtained after comprehensive calculation based on media coverage, social platform discussion heat (such as Weibo topic reading volume, news reprint times), search engine index, etc. Specifically, multiple public opinion events within the same time period can be weighted and summed to further obtain the above-mentioned supply-demand public opinion impact quantification data.

[0043] Exemplarily, taking the pledged goods as steel as an example for illustration, assume the keyword correlation: 0.9 (highly related to "rebars"); event type weight: 0.8 (policy change); timeliness factor: 0.95 (2 days after the event occurs); sentiment tendency value: -0.7 (short-term negative impact on steel prices); spread range data: 0.9 (extremely high discussion heat). At this time, the public opinion event impact weight can be 0.9 * 0.8 * 0.95, approximately equal to 0.684; the supply-demand public opinion impact quantification data can be 0.684 * (-0.7) * 0.9, approximately equal to -0.431. This application does not limit this.

[0044] It should be noted that when constructing the market-goods dual-channel valuation model, a multiple linear regression model, a tree model, a neural network model, etc. can be used as the initial model framework. This application takes the initial model as a multiple linear regression model as an example for illustration, which does not constitute a limitation to this application. The multiple linear regression model can be used to analyze the linear relationship between multiple independent variables and a dependent variable. When constructing the target market-goods dual-channel valuation model, based on the multiple linear regression model as the basic framework, the quantitative data of supply-demand public opinion impact and multi-dimensional visual monitoring data are used as independent variables, and the market value of the pledged goods is used as the dependent variable, and a preliminary valuation model can be established, that is, the reference market-goods dual-channel valuation model.

[0045] The reference market-goods dual-channel valuation model is a preliminarily constructed model, which may have problems such as inaccurate prediction and poor data fitting effect. The training and optimization process can use a large amount of historical data to repeatedly train the reference model to adjust the model parameters and improve the model structure, so as to improve the accuracy and stability of the model, and finally obtain a more accurate target market-goods dual-channel valuation model.

[0046] Specifically, the historical multi-source data of market value, multi-dimensional visual monitoring data, and the corresponding actual market value data of pledged goods can be divided into a training set and a test set. Use the training set data to train the reference market-goods dual-channel valuation model, and continuously adjust the regression coefficients and model structure (such as adding or reducing independent variables, changing the complexity of the model, etc.); then use the test set data to evaluate the trained model, such as calculating metrics such as Mean Squared Error (MSE) and Mean Absolute Error (MAE) to measure the prediction accuracy of the model; further, according to the evaluation results, continuously optimize the model until the model reaches satisfactory performance metrics on the test set. At this time, the obtained model is the target market-goods dual-channel valuation model, which can be used to accurately estimate the market value of the pledged goods.

[0047] Exemplarily, the process of determining the regression coefficients by fitting and calculating the existing data and continuously optimizing the model until the model reaches satisfactory performance metrics on the test set to obtain the target market-goods dual-channel valuation model can be seen in the following formula: Among them, is the current market value data output by the target market-goods dual-channel valuation model; is the market channel weight coefficient, ; is the valuation item for market factors; is the valuation item for the status of the pledged goods; is the public opinion influence coefficient; is the quantified value of the influence of supply and demand public opinion; is the weight corresponding to the market price index; is the market price index; is the weight corresponding to the industry supply and demand index; is the industry supply and demand index; is the weight corresponding to the raw material price fluctuation; is the raw material price fluctuation; is the market random error term; is the weight corresponding to the statistical data of the dynamic quantity of the pledged goods; is the statistical data of the dynamic quantity of the pledged goods; is the weight corresponding to the monitoring data of the volume and shape of the pledged goods; is the monitoring data of the volume and shape of the pledged goods; is the weight corresponding to the estimated data of the dynamic weight of the pledged goods; is the estimated data of the dynamic weight of the pledged goods; is the physical state score of the pledged goods, and the value can be from 0 to 100 points; is the total number of public opinion events; is the index of the public opinion event; is the weight of the i-th public opinion event; is the influence factor of the i-th public opinion event.

[0048] S30: Based on the target market-goods dual-channel valuation model, estimate the market value of the current pledged goods to obtain the current market value data.

[0049] After obtaining the target market-goods dual-channel valuation model, the target market-goods dual-channel valuation model can be used to further perform comprehensive calculations and analyses on the current pledged goods to predict and evaluate the market value of the pledged goods at the current moment.

[0050] The current market value data can be understood as the specific data obtained through market value estimation that can be used to reflect the current market value of the pledged goods. The current market value data can be a specific monetary amount, or in the form of a ratio, index, etc. relative to the initial value. This application does not limit this. It can be understood that the current market value data can provide a quantitative basis for subsequent risk assessment and decision-making.

[0051] S40: When the current market value data meets the preset conditions, generate a risk warning message for the pledged goods.

[0052] Among them, the preset conditions can be understood as threshold conditions or judgment criteria for the current market value data set by the system by default or pre-set manually. Optionally, the preset conditions can be set in advance according to factors such as the risk tolerance of the pledge business and contract agreements; for example, when the market value of the pledged goods drops to a certain percentage (such as a 10% drop relative to the initial valuation), or the pledge rate (the ratio of the value of the pledged goods to the loan balance) is lower than a certain threshold, etc. The present application does not limit this.

[0053] The risk warning information of the pledged goods can be understood as the information generated when the current market value data reaches the preset conditions for the purpose of prompting the existence of risks. The risk warning information of the pledged goods can include but is not limited to the name of the pledged goods, the current market value, the degree of risk, the reason for triggering the warning, etc., for the purpose of reminding relevant personnel to pay attention to the change in the value of the pledged goods and take timely measures to prevent risks.

[0054] Exemplarily, taking the scenario of dynamically monitoring the pledge rate as an example, the ratio of the value of the pledged goods to the loan balance can be calculated in real time, and a warning is triggered when it deviates from the contract threshold. Optionally, a multi-level response mechanism can be set: Level 1 warning (the pledge rate drops by 10%): A risk reminder can be automatically pushed to the financial institution and the goods right holder; Level 2 warning (the pledge rate drops by 20%): Part of the credit line can be frozen and supplementary guarantee is required; Level 3 warning (the pledge rate drops by 30%): The goods disposal plan can be initiated, and the judicial evidence preservation system can be linked to fix the evidence. The present application does not limit this.

[0055] S50: Perform risk warning according to the risk warning information of the pledged goods.

[0056] Performing risk warning can be understood as notifying the relevant parties, such as financial institutions, goods right holders, etc., of the risk warning information of the pledged goods in a specific manner. Optionally, the notification method can be text message, email, system pop-up window, etc. It can be understood that by timely conveying the risk information, it can prompt relevant personnel to respond to the risk status of the pledged goods, such as further evaluating the risk, requiring supplementary guarantee, adjusting the loan strategy, etc., and can reduce the risk of the pledge financing business.

[0057] Furthermore, in the pledged goods management system based on AI vision, there can also be a credible data evidence preservation link, which can ensure the credibility, security and sharing of data through two aspects of blockchain evidence preservation and multi-party collaborative verification.

[0058] Regarding blockchain evidence storage, since blockchain is a distributed ledger technology with characteristics such as decentralization, immutability, and traceability. Therefore, during the management of pledged goods, key data such as goods scanning images and valuation records can be stored on the blockchain in real time. The goods scanning image, as an intuitive record of the physical state of the goods, can reflect information such as the quantity, appearance, and storage status of the goods; the valuation record reflects the value assessment results obtained from the dynamic modeling of the market value of the goods at different time points. When uploading data to the blockchain, the system can use the hash algorithm to encrypt the above data. The hash algorithm can convert data of any length into a hash value of a fixed length, which can be understood as the "digital fingerprint" of the data. Once any change occurs to the data, even if it is just a single-byte change, its hash value will be completely different. The blockchain forms an immutable record chain by concatenating the hash values of the data in chronological order. When it is necessary to verify the integrity of the data, simply recalculate the hash value of the data and compare it with the hash value stored on the chain to determine whether the data has been tampered with. This mechanism can ensure the authenticity and integrity of key data, enabling information such as goods scanning images and valuation records to maintain their originality and credibility throughout the pledge cycle, and providing a reliable basis for subsequent business operations and dispute resolution.

[0059] Regarding multi-party collaborative verification, in order to further enhance the credibility of data and break information barriers, financial institutions, warehousing parties, and regulatory agencies achieve the sharing of trusted data through a consortium blockchain. A consortium blockchain is a blockchain that is only open to specific organizations or groups. In the scenario of pledged goods management, the participating parties can act as nodes of the consortium blockchain and jointly maintain the operation of the blockchain. During their respective business operations, the parties upload the generated data to the consortium blockchain according to established rules. For example, the warehousing party uploads the inbound and outbound records of goods, real-time monitoring data, etc.; the financial institution uploads loan information, pledge rate calculation data; the regulatory agency uploads regulatory instructions and inspection results, etc. After the above data is stored on the chain, it can be viewed and verified by other nodes with permissions. Due to the distributed nature of the blockchain, each node stores the complete or partial ledger data. When a certain node modifies the data, it needs to be recognized by the majority of other nodes to take effect. Therefore, it can greatly increase the difficulty of malicious data tampering. At the same time, the above sharing mode enables all parties to obtain comprehensive information related to the pledged goods in real time, eliminating the information island phenomenon formed due to information non-circulation. Financial institutions can understand the storage status and value changes of goods in real time and better assess risks; the warehousing party can know the loan policies and regulatory requirements of financial institutions and standardize its own operations; the regulatory agency can supervise the entire pledged business process in an all-round way to ensure the legal and compliant conduct of the business. Through multi-party collaborative verification and data sharing, the transparency and efficiency of pledged goods management can be improved, the trust among all parties can be enhanced, and strong guarantee can be provided for the stable development of the pledged financing business.

[0060] It can be seen that in the above solution, based on AI vision technology, multi-dimensional visual monitoring of the current pledged goods is performed to obtain multi-dimensional visual monitoring data, and based on the multi-source data of market value and the multi-dimensional visual monitoring data, a target market-goods dual-channel valuation model is constructed. Thus, based on the target market-goods dual-channel valuation model, the market value of the current pledged goods can be estimated to obtain the current market value data. Further, when the current market value data meets the preset conditions, a risk warning message for the pledged goods is generated. Furthermore, based on the risk warning message for the pledged goods, risk warning can be carried out, and more accurate and effective risk management of the current pledged goods can be achieved.

[0061] Consistent with the above embodiments, please refer to Figure 2 , Figure 2 which is a schematic structural diagram of a terminal provided by an embodiment of the present application, as Figure 2As shown, it includes a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. Among them, the memory is used to store a computer program, and the computer program includes program instructions. The processor is configured to call the program instructions. The above program includes instructions for performing the following steps; Based on AI vision technology, perform multi-dimensional visual monitoring on the current pledged goods to obtain multi-dimensional visual monitoring data; According to multi-source market value data and the multi-dimensional visual monitoring data, construct a target market-goods dual-channel valuation model; Based on the target market-goods dual-channel valuation model, estimate the market value of the current pledged goods to obtain current market value data; When the current market value data meets the preset conditions, generate a risk warning information for the pledged goods; According to the risk warning information for the pledged goods, conduct risk warning.

[0062] In this example, based on AI vision technology, multi-dimensional visual monitoring is performed on the current pledged goods to obtain multi-dimensional visual monitoring data. And according to multi-source market value data and the multi-dimensional visual monitoring data, a target market-goods dual-channel valuation model is constructed. Thus, based on the target market-goods dual-channel valuation model, the market value of the current pledged goods can be estimated to obtain current market value data. Further, when the current market value data meets the preset conditions, a risk warning information for the pledged goods is generated. Furthermore, according to the risk warning information for the pledged goods, risk warning can be conducted, which can realize more accurate and effective risk management of the current pledged goods.

[0063] The above mainly introduces the solution of the embodiment of the present application from the perspective of the execution process on the method side. It can be understood that in order for the terminal to implement the above functions, it includes the corresponding hardware structure and / or software module for executing each function. Those skilled in the art should easily realize that, combining the units and algorithm steps of each example described in the embodiments provided in this article, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the way of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0064] Embodiments of the present application can divide the functions of the terminal according to the above method examples. For example, each functional unit can be divided corresponding to each function, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. It should be noted that the division of units in the embodiments of the present application is illustrative, only a logical function division, and there can be other division methods in actual implementation.

[0065] Consistent with the above, please refer to Figure 3 , Figure 3 which is a schematic structural diagram of a pledged goods management device based on AI vision provided by an embodiment of the present application. As Figure 3 shown, the device includes: A first processing unit 101, configured to perform multi-dimensional visual monitoring on the current pledged goods based on AI vision technology to obtain multi-dimensional visual monitoring data; A second processing unit 102, configured to construct a target market-goods dual-channel valuation model according to multi-source market value data and the multi-dimensional visual monitoring data; A third processing unit 103, configured to estimate the market value of the current pledged goods based on the target market-goods dual-channel valuation model to obtain current market value data; A fourth processing unit 104, configured to generate a pledged goods risk warning message when the current market value data meets a preset condition; A fifth processing unit 105, configured to perform risk warning according to the pledged goods risk warning message.

[0066] In a possible implementation manner, the first processing unit 101, configured to perform multi-dimensional visual monitoring on the current pledged goods based on AI vision technology to obtain multi-dimensional visual monitoring data, is specifically configured to: Collect a current pledged state image of the current pledged goods based on AI vision technology; Perform pledged goods recognition processing on the current pledged state image to obtain pledged goods dynamic quantity statistical data; Obtain current pledged point cloud data of the current pledged goods based on AI vision technology; Adopt point cloud reconstruction technology to perform three-dimensional model construction processing on the current pledged point cloud data to obtain pledged goods volume and shape monitoring data; Construct a volume-weight correlation model based on the pledged goods volume and shape monitoring data, and obtain reference dynamic weight estimation data through the volume-weight correlation model; Extract density-related feature data based on AI vision technology, and perform non-linear correction on the reference dynamic weight estimation data based on the density-related feature data to obtain the dynamic weight estimation data of the pledged goods; Determine multi-dimensional visual monitoring data according to the dynamic quantity statistical data of the pledged goods, the volume and shape monitoring data of the pledged goods, and the dynamic weight estimation data of the pledged goods.

[0067] In a possible implementation, the second processing unit 102 is configured to construct a target market-goods dual-channel valuation model according to the multi-source market value data and the multi-dimensional visual monitoring data, specifically: Preprocess the multi-source market value data to obtain target multi-source market value data; Conduct an impact assessment of supply and demand public opinion based on the target multi-source market value data to obtain quantified supply and demand public opinion impact data; Based on the initial model, obtain a reference market-goods dual-channel valuation model according to the quantified supply and demand public opinion impact data and the multi-dimensional visual monitoring data; Train and optimize the reference market-goods dual-channel valuation model to obtain the target market-goods dual-channel valuation model.

[0068] In a possible implementation, the second processing unit 102 is configured to conduct an impact assessment of supply and demand public opinion based on the target multi-source market value data to obtain quantified supply and demand public opinion impact data, specifically: Establish and obtain a keyword library for pledged goods according to the target multi-source market value data for different types of pledged goods; Judge the sentiment tendency of public opinion events for the target multi-source market value data to obtain sentiment tendency impact data; Conduct impact analysis on each public opinion event according to the keyword library of pledged goods to obtain the impact weight of public opinion events; Quantify the value impact of each public opinion event on the pledged goods according to the sentiment tendency impact data, the impact weight of public opinion events, and the spread range data of each public opinion event to obtain quantified supply and demand public opinion impact data.

[0069] In a possible implementation, the second processing unit 102 is configured to conduct impact analysis on each public opinion event according to the keyword library of pledged goods to obtain the impact weight of public opinion events, specifically: Evaluate the relevance of each public opinion event according to the keyword library of pledged goods to obtain the keyword relevance; Calculate the weight of the event type of each public opinion event according to the keyword library of pledged goods to obtain the weight of the public opinion event type; Conduct a timeliness assessment for each public opinion event to obtain a timeliness factor; Determine the influence weight of the public opinion event according to the keyword relevance, the weight of the public opinion event type, and the timeliness factor.

[0070] The embodiments of the present application also provide a computer storage medium. The computer storage medium stores a computer program for electronic data exchange, and the computer program enables a computer to execute some or all of the steps of any one of the pledge goods management methods based on AI vision described in the foregoing method embodiments.

[0071] The embodiments of the present application also provide a computer program product. The computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program enables a computer to execute some or all of the steps of any one of the pledge goods management methods based on AI vision described in the foregoing method embodiments.

[0072] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0073] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0074] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical or other form.

[0075] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0076] In addition, in each of the embodiments of the present application, each functional unit may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software program module.

[0077] If the above-mentioned integrated unit is implemented in the form of a software program module and sold or used as an independent product, it may be stored in a computer-readable memory. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned memory includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs, etc., which are various media that can store program codes.

[0078] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program. This program can be stored in a computer-readable memory, and the memory may include: flash drives, read-only memories, random access memories, magnetic disks, or optical discs, etc.

[0079] The above has introduced the embodiments of the present application in detail. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for managing pledged goods based on AI vision, characterized in that, The method includes: Based on AI vision technology, conduct multi-dimensional visual monitoring on the current pledged goods to obtain multi-dimensional visual monitoring data; Construct a target market-goods dual-channel valuation model according to the multi-source market value data and the multi-dimensional visual monitoring data; Based on the target market-goods dual-channel valuation model, estimate the market value of the current pledged goods to obtain the current market value data; When the current market value data meets the preset conditions, generate a risk warning message for the pledged goods; Conduct risk warning according to the risk warning message of the pledged goods.

2. The method for managing pledged goods based on AI vision according to claim 1, wherein, The step of, based on AI vision technology, conducting multi-dimensional visual monitoring on the current pledged goods to obtain multi-dimensional visual monitoring data includes: Based on AI vision technology, collect the current pledged state image of the current pledged goods; Perform identification processing on the current pledged state image of the pledged goods to obtain the dynamic quantity statistical data of the pledged goods; Based on AI vision technology, obtain the current point cloud data of the current pledged goods; Adopt point cloud reconstruction technology to conduct three-dimensional model construction processing on the current point cloud data of the pledged goods to obtain the volume and shape monitoring data of the pledged goods; Construct a volume-weight correlation model based on the volume and shape monitoring data of the pledged goods, and obtain the reference dynamic weight estimation data through the volume-weight correlation model; Extract density-related feature data based on AI vision technology, and perform non-linear correction on the reference dynamic weight estimation data based on the density-related feature data to obtain the dynamic weight estimation data of the pledged goods; Determine the multi-dimensional visual monitoring data according to the dynamic quantity statistical data of the pledged goods, the volume and shape monitoring data of the pledged goods, and the dynamic weight estimation data of the pledged goods.

3. The method for managing pledged goods based on AI vision according to claim 2, wherein The step of constructing a target market-goods dual-channel valuation model according to the multi-source market value data and the multi-dimensional visual monitoring data includes: Preprocess the multi-source market value data to obtain the target multi-source market value data; Conduct supply-demand public opinion impact assessment according to the target multi-source market value data to obtain the quantified supply-demand public opinion impact data; Based on the initial model, obtain a reference market-goods dual-channel valuation model according to the quantified supply-demand public opinion impact data and the multi-dimensional visual monitoring data; Train and optimize the reference market-goods dual-channel valuation model to obtain the target market-goods dual-channel valuation model.

4. The method for managing pledged goods based on AI vision according to claim 3, wherein, The step of conducting supply-demand public opinion impact assessment according to the target multi-source market value data to obtain the quantified supply-demand public opinion impact data includes: According to the target multi-source market value data, establish and obtain a keyword library for pledged goods for different types of pledged goods; Judge the sentiment tendency of public opinion events for the target multi-source market value data to obtain the sentiment tendency impact data; Conduct impact analysis on each public opinion event according to the keyword library of the pledged goods to obtain the impact weight of the public opinion event; Quantify the value impact of each public opinion event on the pledged goods according to the sentiment tendency impact data, the impact weight of the public opinion event, and the spread range data of each public opinion event to obtain the quantified supply-demand public opinion impact data.

5. The method for managing pledged goods based on AI vision according to claim 4, wherein Performing impact analysis on each public opinion event according to the pledged goods keyword library to obtain the impact weight of the public opinion event, including: Evaluating the correlation degree of each public opinion event according to the pledged goods keyword library to obtain the keyword correlation degree; Calculating the weight of the event type of each public opinion event according to the pledged goods keyword library to obtain the weight of the public opinion event type; Evaluating the timeliness of each public opinion event to obtain the timeliness factor; Determining the impact weight of the public opinion event according to the keyword correlation degree, the weight of the public opinion event type and the timeliness factor.

6. A pledge goods management device based on AI vision, characterized in that, The device includes: The first processing unit is used to perform multi-dimensional visual monitoring on the current pledged goods based on AI vision technology to obtain multi-dimensional visual monitoring data; The second processing unit is used to construct a target market-goods dual-channel valuation model according to the multi-source data of the market value and the multi-dimensional visual monitoring data; The third processing unit is used to estimate the market value of the current pledged goods based on the target market-goods dual-channel valuation model to obtain the current market value data; The fourth processing unit is used to generate a risk warning information for the pledged goods when the current market value data meets the preset conditions; The fifth processing unit is used to perform risk warning according to the risk warning information of the pledged goods.

7. The AI vision-based pledged goods management device according to claim 6, characterized in that, The first processing unit is used to perform multi-dimensional visual monitoring on the current pledged goods based on AI vision technology to obtain multi-dimensional visual monitoring data, specifically: Collecting the current pledged state image of the current pledged goods based on AI vision technology; Performing identification processing on the current pledged state image of the pledged goods to obtain the dynamic quantity statistical data of the pledged goods; Obtaining the current pledged point cloud data of the current pledged goods based on AI vision technology; Using point cloud reconstruction technology to perform three-dimensional model construction processing on the current pledged point cloud data to obtain the volume and shape monitoring data of the pledged goods; Constructing a volume-weight correlation model based on the volume and shape monitoring data of the pledged goods, and obtaining reference dynamic weight estimation data through the volume-weight correlation model; Extracting density-related feature data based on AI vision technology, and performing non-linear correction on the reference dynamic weight estimation data based on the density-related feature data to obtain the dynamic weight estimation data of the pledged goods; Determining the multi-dimensional visual monitoring data according to the dynamic quantity statistical data of the pledged goods, the volume and shape monitoring data of the pledged goods and the dynamic weight estimation data of the pledged goods.

8. The AI vision-based pledged goods management device according to claim 7, characterized in that The second processing unit is used to construct a target market-goods dual-channel valuation model according to the multi-source data of the market value and the multi-dimensional visual monitoring data, specifically: Performing preprocessing on the multi-source data of the market value to obtain target market value multi-source data; Performing impact evaluation of supply and demand public opinion according to the target market value multi-source data to obtain quantitative data of the impact of supply and demand public opinion; Based on the initial model, obtaining a reference market-goods dual-channel valuation model according to the quantitative data of the impact of supply and demand public opinion and the multi-dimensional visual monitoring data; Train and optimize the reference market-goods dual-channel valuation model to obtain the target market-goods dual-channel valuation model.

9. A terminal, characterized in that, It includes a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. Among them, the memory is used to store computer programs, and the computer programs include program instructions. The processor is configured to call the program instructions to execute the method according to any one of claims 1-5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by the processor, the processor is caused to execute the method according to any one of claims 1-5.

Images