Search

CN-122024061-A - Book asset processing method and device based on artificial intelligence

CN122024061ACN 122024061 ACN122024061 ACN 122024061ACN-122024061-A

Abstract

The invention relates to the technical field of intelligent library management and service, and discloses a book asset processing method and device based on artificial intelligence, comprising the steps of acquiring multispectral imaging data of book assets to be detected, extracting physical defect characteristics from the multispectral imaging data, and generating physical defect characteristic vectors; the method comprises the steps of inputting the physical defect feature vector into a density clustering algorithm to perform clustering processing to generate a defect mode cluster, obtaining supply chain metadata of books in the defect mode cluster, enabling nodes in a supply chain network diagram to represent supply chain entities and edges to represent supply relations, identifying intersection nodes which occur in multiple supply chain paths simultaneously by using a graph algorithm, marking the nodes with the intersection frequency exceeding a threshold value as suspected defect source nodes, traversing downstream supply paths of the suspected defect source nodes, and generating a risk book early warning list. The invention realizes the diffusion identification from the found defects to the potential defects, and protects the book assets without problems in advance.

Inventors

  • CHEN XIAOYU

Assignees

  • 广东工贸职业技术学院

Dates

Publication Date
20260512
Application Date
20260202

Claims (10)

  1. 1. The book asset processing method based on artificial intelligence is characterized by comprising the following steps of: acquiring multispectral imaging data of a book asset to be detected, extracting physical defect characteristics from the multispectral imaging data, and generating a physical defect characteristic vector; inputting the physical defect feature vector into a density clustering algorithm for clustering, identifying book groups with similar physical degradation modes, and generating a defect mode cluster; acquiring the supply chain metadata of each book in the defect mode cluster, and generating a supply chain network diagram based on the supply chain metadata, wherein nodes in the supply chain network diagram represent supply chain entities, and edges represent supply relations; extracting supply chain paths corresponding to books in the defect mode cluster from the supply chain network diagram, identifying intersection nodes which appear in a plurality of supply chain paths simultaneously by using a graph algorithm, and marking the nodes with the intersection frequency exceeding a threshold value as suspected defect source nodes; and traversing the downstream supply path based on the suspected defect source node, extracting an affected downstream book list, and generating a risk book early warning list.
  2. 2. The method of claim 1, wherein the multispectral imaging data comprises an ultraviolet reflectance image, a visible light reflectance image, and a near infrared reflectance image, wherein the physical defect characteristics comprise paper colorimetric parameters, fiber morphology characteristics, carbonyl content indicators, and cellulose polymerization indicators, and wherein the ultraviolet reflectance image is used to detect autofluorescence characteristics of mold metabolites.
  3. 3. The method of claim 1, wherein inputting the physical defect feature vector into a density clustering algorithm for clustering, comprises: calculating the reflectivity ratio and the spectral gradient of each wave band in the physical defect feature vector, extracting fingerprint features of chemical components of paper, and generating a defect fingerprint data set; And inputting the defect fingerprint data set into a DBSCAN density clustering algorithm, identifying book sample points which are mutually adjacent in a feature space according to a neighborhood radius parameter and a minimum sample number parameter, dividing the mutually adjacent sample points into the same cluster, and generating a defect mode cluster set.
  4. 4. The method of claim 1, wherein the supply chain metadata includes a press identification, a print job identification, a paper supplier identification, and a raw material lot number, wherein the supply chain network graph is a directed multi-level network graph, and wherein each directed edge represents a material supply relationship from an upstream node to a downstream node.
  5. 5. The method of claim 1, wherein identifying intersection nodes that occur simultaneously in multiple supply chain paths using a graph algorithm comprises: Extracting a supply chain path set corresponding to each book in the defect mode cluster, wherein each supply chain path is a directed node sequence from a raw material provider to a final book; Analyzing the supply chain path set by using a maximum public sub-graph algorithm, and counting the occurrence frequency of each node in different paths; Marking the nodes with occurrence frequency exceeding a preset threshold as suspected defect source nodes, and generating a suspected defect source node set.
  6. 6. The method of claim 1, further comprising, prior to identifying the suspected defect source node: Acquiring credit rating data, financial condition data and historical quality complaint records of each enterprise node in the supply chain network diagram; inputting the credit rating data, the financial condition data and the historical quality complaint records as features into a logistic regression algorithm, and calculating quality risk scores of all nodes; Marking the nodes with the quality risk scores higher than a preset threshold as high-risk nodes, and generating node risk labels; and when the suspected defect source nodes are identified, the nodes meeting the high-intersection frequency and high-risk labels are incorporated into a suspected defect source node set.
  7. 7. The method of claim 6, further comprising, after generating the set of suspected defect source nodes: Counting the proportion of defects in the downstream books of each node in the suspected defect source node set as likelihood probability; The node risk scores are used as prior probabilities, a Bayesian inference algorithm is utilized to calculate posterior probability of each node as a real defect source, and the posterior probability is the sum of the products of likelihood probability and prior probability divided by the products of likelihood probability and prior probability of all suspected nodes; and sorting from high to low according to the posterior probability, and generating a defect source credibility sorting list.
  8. 8. The method of claim 7, further comprising, after generating the defect source reliability ranking list: The provider nodes with posterior probability exceeding a preset threshold in the defect source credibility sequencing list are included in a provider quality blacklist; querying the supply chain network diagram for alternative nodes having the same supply function as the suppliers in the supplier quality black list; And acquiring historical quality data and cost data of the substitute nodes, calculating substitute cost and quality reliability indexes, and generating a provider switching scheme.
  9. 9. The method of claim 1, wherein generating a risk book pre-warning list comprises: acquiring all downstream supply paths of the suspected defect source node; Traversing the supply chain network diagram along the direction of the directed edge from the suspected defect source node by using a graph traversing algorithm, and extracting all downstream book lists influenced by the suspected defect source node; Aiming at books which do not show defects in the downstream book list, calculating defect latency probability according to the similarity between the books and the supply chain path of the books which have defects; and generating a risk book early warning list based on the defect latency probability, wherein the risk book early warning list comprises book identifications and corresponding risk grades.
  10. 10. A book asset processing system based on artificial intelligence for performing the method of any one of claims 1-9, comprising: The defect feature extraction module is used for acquiring multispectral imaging data of the book asset to be detected, extracting physical defect features from the multispectral imaging data and generating a physical defect feature vector; the pattern clustering module is used for inputting the physical defect feature vector into a density clustering algorithm to perform clustering processing, identifying book groups with similar physical degradation patterns and generating defect pattern clusters; The supply chain network generation module is used for acquiring the supply chain metadata of each book in the defect mode cluster and generating a supply chain network diagram based on the supply chain metadata; the defect source identification module is used for extracting supply chain paths of the defect books from the supply chain network diagram, identifying intersection nodes which appear in a plurality of supply chain paths simultaneously by using a graph algorithm, and generating a suspected defect source node set; and the risk early warning module is used for generating a risk book early warning list based on the suspected defect source node traversing the downstream supply path.

Description

Book asset processing method and device based on artificial intelligence Technical Field The invention relates to the technical field of intelligent library management and service, in particular to a book asset processing method and device based on artificial intelligence. Background The invention relates to the technical field of book asset processing, in particular to a book asset quality defect tracing processing method based on artificial intelligence. In the practice of managing book assets in a library, a publisher, or the like, a supply chain of book assets involves a plurality of levels of nodes such as publishers, printing factories, paper suppliers, raw material suppliers, and the like. When book assets have quality defects such as yellowing, mildew, pH anomalies, etc., these defects may originate from a raw material defect or a process problem at a node upstream of the supply chain. Since the same upstream provider may supply multiple publishers through different paths, its quality defects may be presented in a similar fashion among multiple seemingly unrelated book assets. The conventional book asset processing method has the technical problems that firstly, the conventional manual visual inspection can only identify visible mildew spots on the surface of books, chemical component changes and early degradation signals in paper cannot be detected, deep physical defect features are difficult to extract, secondly, books with the problems are treated as isolated events, common degradation modes of batch book assets cannot be identified, systematic quality problems are omitted, and thirdly, the conventional method lacks the capability of reversely tracing and sharing an upstream provider from a defective book due to the multistage and complexity of a book supply chain, and the root cause of quality defects cannot be located. These problems result in continuous supply by the problem provider, resulting in systematic diffusion and batch devaluation of book asset quality risks. The hardware environment comprises a book detection workstation provided with a multispectral imaging device, a Raman spectrometer and a server, wherein the book detection workstation is provided with the multispectral imaging device which can collect reflectivity images of ultraviolet, visible light and near infrared bands, the Raman spectrometer is provided with the Raman spectrometer and is used for collecting molecular vibration spectrum data of paper, and the server is provided with a graph database and is used for storing and processing supply chain network data. Disclosure of Invention The invention provides a book asset processing method and device based on artificial intelligence, which solve the technical problems that the traditional method in the related technology cannot trace back the common upstream suppliers from the defective books in a reverse direction, and the traditional method cannot detect deep physical defects. The invention provides a book asset processing method based on artificial intelligence, which comprises the following steps: acquiring multispectral imaging data of a book asset to be detected, extracting physical defect characteristics from the multispectral imaging data, and generating a physical defect characteristic vector; inputting the physical defect feature vector into a density clustering algorithm for clustering, identifying book groups with similar physical degradation modes, and generating a defect mode cluster; acquiring the supply chain metadata of each book in the defect mode cluster, and generating a supply chain network diagram based on the supply chain metadata, wherein nodes in the supply chain network diagram represent supply chain entities, and edges represent supply relations; extracting supply chain paths corresponding to books in the defect mode cluster from the supply chain network diagram, identifying intersection nodes which appear in a plurality of supply chain paths simultaneously by using a graph algorithm, and marking the nodes with the intersection frequency exceeding a threshold value as suspected defect source nodes; and traversing the downstream supply path based on the suspected defect source node, extracting an affected downstream book list, and generating a risk book early warning list. Further, the multispectral imaging data comprises an ultraviolet band reflectivity image, a visible light band reflectivity image and a near infrared band reflectivity image, the physical defect characteristics comprise paper chromaticity parameters, fiber morphological characteristics, carbonyl content indexes and cellulose polymerization degree indexes, and the ultraviolet band reflectivity image is used for detecting autofluorescence characteristics of mold metabolites. Further, inputting the physical defect feature vector into a density clustering algorithm for clustering processing, including: calculating the reflectivity ratio and the spectral gradient of each wave band in