Method and system of verifying authenticity of declaration information, device and medium

Patent metadata

Field	Value
Publication number	US-12488575-B2
Application number	US-202217980831-A
Country	US
Kind code	B2
Filing date	Nov 4, 2022
Priority date	Nov 5, 2021
Publication date	Dec 2, 2025
Grant date	Dec 2, 2025

Abstract

Provided is a method of verifying an authenticity of declaration information. The method includes: acquiring a machine-detected radiation image obtained by scanning a container loaded with an article; acquiring a declaration information for declaring the article in the container; performing an identification on an image information of the article in the machine-detected radiation image to obtain an image feature corresponding to the machine-detected radiation image; performing an identification on a text information of the article in the declaration information to obtain a text feature corresponding to the declaration information; screening a declaration category of the article in the container by taking the image feature as an input information and the text feature as an external introduction feature; and determining that the declaration information is in doubt when a declaration category of at least one article in the container does not belong to a declaration category in the declaration information.

First claim

What is claimed is:

1 . A method of verifying an authenticity of declaration information, comprising:
acquiring a machine-detected radiation image obtained by scanning a container loaded with an article; acquiring a declaration information for declaring the article in the container; performing an identification on an image information of the article in the machine-detected radiation image to obtain an image feature corresponding to the machine-detected radiation image; performing an identification on a text information of the article in the declaration information to obtain a text feature corresponding to the declaration information, wherein the text feature is configured to represent a declaration category to which the article in the declaration information belongs; screening a declaration category of the article in the container by taking the image feature as an input information and the text feature as an external introduction feature; and determining that the declaration information is in doubt when a declaration category of at least one article in the container does not belong to a declaration category in the declaration information, wherein the text feature comprises text feature vectors corresponding to N2 articles in the declaration information, respectively, wherein N2 is an integer greater than or equal to 1, wherein the performing an identification on a text information of the article in the declaration information to obtain a text feature corresponding to the declaration information comprises: extracting a name information and a specification model information of each article in the declaration information; for each article, processing the name information into a first statement, and processing the specification model information into a second statement; determining the first statement and the second statement corresponding to a same article as an input of a text feature extraction projection module, and classifying the declaration category to which the article belongs using the text feature extraction projection module; and determining an output result of the text feature extraction projection module for each article category as the text feature vector corresponding to the article; wherein N2 text feature vectors are correspondingly obtained for N2 articles.

2 . The method according to claim 1 , wherein the image feature comprises N1 first image feature vectors corresponding to image information of N1 articles in the machine-detected radiation image, respectively, wherein N1 is an integer greater than or equal to 1.

3 . The method according to claim 2 , wherein the performing an identification on an image information of the article in the machine-detected radiation image to obtain an image feature corresponding to the machine-detected radiation image comprises:
dividing different articles in the machine-detected radiation image into independent image blocks by using a target detection algorithm, to obtain N1 image blocks; extracting a second image feature vector corresponding to each of the image blocks; and obtaining, based on the second image feature vector corresponding to each of the image blocks, the first image feature vector corresponding to the image information of the article represented by the image block.

4 . The method according to claim 3 , wherein the extracting a second image feature vector corresponding to each of the image blocks comprises:
performing an image identification on each of the image blocks by using an image feature extraction module, to obtain the second image feature vector corresponding to each of the image blocks, wherein the image feature extraction module comprises a convolution neural network.

5 . The method according to claim 4 , wherein the image feature extraction module comprises a network structure taking resnet as a basic network and adding SE-block after a resnet pooling layer.

6 . The method according to claim 3 , wherein the obtaining the first image feature vector corresponding to the image information of the article represented by the image block based on the second image feature vector corresponding to each of the image blocks comprises:
acquiring a position information of each of the image blocks in the machine-detected radiation image; and obtaining the first image feature vector based on the second image feature vector and the position information corresponding to a same image block.

7 . The method according to claim 6 , wherein the obtaining the first image feature vector based on the second image feature vector and the position information corresponding to a same image block comprises:
processing the second image feature vector using the position information of the image block; inputting the processed second image feature vector into an encoder; and obtaining an output of the encoder to obtain the first image feature vector corresponding to the image block.

8 . The method according to claim 7 , wherein the screening a declaration category of the article in the container by taking the image feature as an input information and the text feature as an external introduction feature comprises:
screening the declaration category of the article in the container using a cross-modal decoder by taking the image feature as an input information of the cross-modal decoder and the text feature as an external introduction feature of an attention mechanism of the cross-modal decoder.

9 . The method according to claim 8 , wherein the encoder and the cross-modal decoder are jointly trained.

10 . The method according to claim 9 , wherein the encoder adopts a transformer encoder model.

11 . The method according to claim 10 , wherein the cross-modal decoder adopts a transformer decoder model.

12 . The method according to claim 1 , wherein the text feature extraction projection module adopts a BERT model.

13 . A system of verifying an authenticity of declaration information, comprising:
an information acquisition subsystem configured to acquire a machine-detected radiation image obtained by scanning a container loaded with an article, and acquire a declaration information for declaring the article in the container; a feature extraction subsystem configured to perform an identification on an image information of the article in the machine-detected radiation image to obtain an image feature corresponding to the machine-detected radiation image, and perform an identification on a text information of the article in the declaration information to obtain a text feature corresponding to the declaration information, wherein, the text feature is configured to represent a declaration category to which the article in the declaration information belongs; a feature fusion subsystem configured to screen a declaration category of the article in the container by taking the image feature as an input information and the text feature as an external introduction feature; and a conclusion determination subsystem configured to determine that the declaration information is in doubt when a declaration category of at least one article in the container does not belong to a declaration category in the declaration information, wherein the text feature comprises text feature vectors corresponding to N2 articles in the declaration information, respectively, wherein N2 is an integer greater than or equal to 1, wherein the performing an identification on a text information of the article in the declaration information to obtain a text feature corresponding to the declaration information comprises: extracting a name information and a specification model information of each article in the d

Assignees

• Nuctech Co Ltd
• Univ Tsinghua

Inventors

• Zhang Li
• Xing Yuxiang
• Tang Hu
• Sun Yunda
• Fu Gang
• Li Qiang
• Zhang Yan
• Wu Wubin

Classifications

• G06V10/26

  Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion · CPC title

• G06V20/62

  Text, e.g. of license plates, overlay texts or captions on TV images · CPC title

• G06V10/82

  using neural networks · CPC title

• G06F40/295

  Named entity recognition · CPC title

• G06N3/08

  Learning methods · CPC title