Search

CN-121980049-A - Logistics mail identification method and device, electronic equipment and storage medium

CN121980049ACN 121980049 ACN121980049 ACN 121980049ACN-121980049-A

Abstract

The embodiment of the application provides a method and a device for identifying a logistics support, electronic equipment and a storage medium, and belongs to the technical field of logistics. The method comprises the steps of constructing a support object label database based on object types in a historical logistics order, obtaining a public image set, extracting object text entities of each public image in the public image set to obtain object names, expanding the support object label database according to the object names and the public images corresponding to the object names, training a preset initial recognition model according to the expanded support object label database to obtain a target recognition model, obtaining target support object images to be recognized, and calling the target recognition model to carry out support object recognition on the target support object images to obtain support object recognition results. The embodiment of the application can improve the accuracy of the identification of the mail holding object.

Inventors

  • Pu Anni

Assignees

  • 顺丰科技有限公司

Dates

Publication Date
20260505
Application Date
20251229

Claims (10)

  1. 1. A method for identifying a logistic mail, the method comprising: constructing a consignment tag database based on the object types in the historical logistics order, wherein the consignment tag database comprises a plurality of consignment tags and associated images of each consignment tag, and the associated images are object images acquired in a logistics link; Acquiring a public image set, and extracting an object text entity for each public image in the public image set to obtain an object name; Expanding the consignment tag database according to the article name and the public image corresponding to the article name; Training a preset initial recognition model according to the expanded support object label database to obtain a target recognition model; And acquiring a target consignment image to be identified, and calling the target identification model to carry out consignment identification on the target consignment image to obtain a consignment identification result, wherein the consignment identification result comprises a consignment label of the target consignment image or a label indicating that consignment cannot be carried out.
  2. 2. The method of claim 1, wherein augmenting the hosted object tag database based on the item name and the public image corresponding to the item name comprises: performing label expansion on the carrier label database according to the object names, wherein the object names are used as fine carrier labels in the carrier label database, and the object categories are used as generalized carrier labels in the carrier label database; classifying the fine support object labels, and carrying out label association on the fine support object labels and the summarized support object labels with the same category according to classification results; and carrying out associated image expansion on the summarized support object labels with the label association relation according to the public images corresponding to the fine support object labels in the public image set.
  3. 3. The method of claim 2, wherein augmenting the hosted object tag database based on the item name and the public image corresponding to the item name further comprises: invoking a search engine to search the image of the fine mail label; and carrying out associated image expansion on the fine mail label according to the searched image.
  4. 4. The method of claim 1, wherein training the preset initial recognition model according to the expanded host tag database to obtain the target recognition model comprises: Aiming at the expanded support object label in the support object label database, calling a preset text description model to carry out text description based on the support object label, an associated image of the support object label and a preset description prompt word, and obtaining a reference description text; continuously pre-training a preset initial recognition model according to the associated image of the mail supporting object label and the corresponding reference description text; and generating a prompt word, a support object label and an associated image of the support object label according to the preset first label, and performing fine tuning training on the initial recognition model after the continuous pre-training to obtain a target recognition model.
  5. 5. The method of claim 1, wherein before training the pre-set initial recognition model according to the expanded host tag database to obtain the target recognition model, the method further comprises: Classifying the appearance of the article according to the associated image of each consignment label in the consignment label database; and filtering the associated images of the outer package represented by the article appearance classification result.
  6. 6. The method of claim 1, wherein before training the pre-set initial recognition model according to the expanded host tag database to obtain the target recognition model, the method further comprises: converting all the consignment labels in the consignment label database into label vectors, and converting all the associated images in the consignment label database into image vectors; For each consignment label, carrying out vector matching on a label vector corresponding to the consignment label and all image vectors, and determining the image vector corresponding to the associated image of the consignment label as a seed vector when the number of the matched image vectors is smaller than a preset number threshold; and carrying out similarity search in all non-seed vectors according to the seed vectors, and carrying out associated image expansion on the carrier labels according to the similarity search result.
  7. 7. The method of claim 6, wherein the performing associated image augmentation on the host tag based on the similarity search results comprises: determining an associated image corresponding to a non-seed vector with the seed vector similarity greater than a preset first similarity threshold as a candidate image; Invoking a preset label generation model to generate labels for the candidate images based on a preset second label generation prompt word to obtain candidate labels; and when the similarity between the candidate label and the corresponding mail holding label is larger than a preset second similarity threshold, carrying out associated image expansion on the corresponding mail holding label according to the candidate image.
  8. 8. A device for identifying a physical distribution mailpiece, the device comprising: The database construction unit is used for constructing a supporting object label database based on the object types in the historical logistics orders, wherein the supporting object label database comprises a plurality of supporting object labels and associated images of each supporting object label, and the associated images are object images acquired in a logistics link; The name acquisition unit is used for acquiring a public image set, and extracting an object text entity from each public image in the public image set to obtain an object name; the database expansion unit is used for expanding the support object label database according to the object names and the public images corresponding to the object names; the model training unit is used for training a preset initial recognition model according to the expanded support object label database to obtain a target recognition model; the device comprises a supporting object identification unit, a target supporting object identification model and a supporting object identification unit, wherein the supporting object identification unit is used for acquiring a target supporting object image to be identified, and calling the target identification model to carry out supporting object identification on the target supporting object image to obtain a supporting object identification result, and the supporting object identification result comprises a supporting object label of the target supporting object image or a label indicating that supporting is impossible.
  9. 9. An electronic device comprising a memory storing a computer program and a processor implementing the method of any of claims 1 to 7 when the computer program is executed by the processor.
  10. 10. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 7.

Description

Logistics mail identification method and device, electronic equipment and storage medium Technical Field The present application relates to the field of logistics technologies, and in particular, to a method and apparatus for identifying a logistic mail, an electronic device, and a storage medium. Background In the logistics industry, accurate identification of articles (i.e., shipping articles) shipped by customers is important for links such as compliance security check, package recommendation, freight calculation, and the like. In the related art, the carrying object identification is carried out by relying on manual judgment or a simple classification model, but the accuracy of carrying object identification in a logistics scene by the simple classification model is lower. Disclosure of Invention The embodiment of the application mainly aims to provide a method and a device for identifying logistic mail, electronic equipment and a storage medium, which can improve the accuracy of mail identification. In order to achieve the above object, a first aspect of an embodiment of the present application provides a method for identifying a logistic mail, where the method includes: constructing a consignment tag database based on the object types in the historical logistics order, wherein the consignment tag database comprises a plurality of consignment tags and associated images of each consignment tag, and the associated images are object images acquired by a logistics link; Acquiring a public image set, and extracting an object text entity for each public image in the public image set to obtain an object name; Expanding the consignment tag database according to the article name and the public image corresponding to the article name; Training a preset initial recognition model according to the expanded support object label database to obtain a target recognition model; And acquiring a target consignment image to be identified, and calling the target identification model to carry out consignment identification on the target consignment image to obtain a consignment identification result, wherein the consignment identification result comprises a consignment label of the target consignment image or a label indicating that consignment cannot be carried out. In order to achieve the above object, a second aspect of the embodiments of the present application provides a device for identifying a logistic mail, including: The database construction unit is used for constructing a supporting object label database based on the object types in the historical logistics orders, wherein the supporting object label database comprises a plurality of supporting object labels and associated images of each supporting object label, and the associated images are object images acquired in a logistics link; The name acquisition unit is used for acquiring a public image set, and extracting an object text entity from each public image in the public image set to obtain an object name; the database expansion unit is used for expanding the support object label database according to the object names and the public images corresponding to the object names; the model training unit is used for training a preset initial recognition model according to the expanded support object label database to obtain a target recognition model; the device comprises a supporting object identification unit, a target supporting object identification model and a supporting object identification unit, wherein the supporting object identification unit is used for acquiring a target supporting object image to be identified, and calling the target identification model to carry out supporting object identification on the target supporting object image to obtain a supporting object identification result, and the supporting object identification result comprises a supporting object label of the target supporting object image or a label indicating that supporting is impossible. To achieve the above object, a third aspect of the embodiments of the present application provides an electronic device, where the electronic device includes a memory and a processor, and the memory stores a computer program, and the processor executes the computer program to implement a method according to any one of the embodiments of the first aspect. To achieve the above object, a fourth aspect of the embodiments of the present application further proposes a computer-readable storage medium storing a computer program, which when executed by a processor, implements the method according to any one of the embodiments of the first aspect. According to the logistics support object identification method and device, the electronic equipment and the storage medium, the support object label database is built through the historical logistics orders, and the support object label database is expanded through the object names obtained based on the public image sets and the public images corresponding to the