Convolutional Neural Network Item Authentication System

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing methods for authenticating luxury goods are time-consuming, costly, and lack accuracy, with no standardized professional qualification or formal training, leading to subjective judgments and potential black market interests, while current deep learning technologies have not effectively addressed the need for quick and reliable authentication.

Innovation Solution

An item authentication method using pre-processing techniques such as size modification and feature approximation degree clustering, combined with convolutional neural networks like VGG19, RESNET54, and WRESNET16, to identify target authentication points and determine authenticity through a decision tree algorithm, setting thresholds for authentication results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If manual authentication is used for luxury goods, then authentication can be performed with basic knowledge, but authentication accuracy is low and results are subjective

Engineering Contradiction:
Improveauthentication accessibilityVSAvoidauthentication accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The patent replaces manual mechanical authentication with an automated image processing system using convolutional neural networks. The system automatically extracts features from images of luxury goods and performs authentication without human intervention, thereby eliminating subjectivity while maintaining ease of operation through automated workflows.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces an image-based intermediate representation as a mediator between the luxury good and the authentication decision. Instead of direct human inspection, the system uses processed images and extracted features as an intermediary that enables objective, consistent authentication while remaining accessible to users through simple image uploads.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If multiple deep learning models are used for authentication, then authentication accuracy is improved, but system complexity increases

Engineering Contradiction:
Improveauthentication accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines multiple convolutional neural network models (VGG19, RESNET54, WRESNET16) into a unified authentication system. The models work together to extract complementary features from images, and their results are integrated through a decision tree algorithm, achieving higher accuracy while managing complexity through systematic integration.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent divides the authentication task into separate specialized models, each trained to extract specific types of features from images. This segmentation allows each model to focus on particular authentication aspects, improving overall accuracy while enabling modular system architecture that manages complexity through division of labor.

Inventive Principle:
Principle #1Segmentation

3Productivity

If image-based authentication is implemented, then authentication speed is improved, but authentication reliability may be compromised without professional oversight

Engineering Contradiction:
Improveauthentication speedVSAvoidauthentication trustworthiness
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent implements a self-service authentication system where the automated model performs authentication without requiring professional human oversight. The system independently processes images, extracts features, and generates authentication results, achieving both high speed and reliability through sophisticated algorithmic decision-making that eliminates human error and bias.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3627392B1Object identification method, system and device, and storage medium
Publication Date: 2024.09.11 TURING AI INST NANJING CO LTD
  • EP3627392B1 patent drawingFigure 1~2
  • EP3627392B1 patent drawingFigure 3~5
  • EP3627392B1 patent drawingFigure 6~7

AI summary

The application provides an item authentication method, system, device and storage medium. The method comprises: acquiring a plurality of images taken on an item to be authenticated, each of the images contains at least one target authentication point; acquiring a single-point score corresponding to each of a plurality of target authentication points through identifying the plurality of images by a plurality of trained convolutional neural network models respectively, each of the trained convolutional neural network models is corresponding to the at least one target authentication point; obtaining a total score through performing a weighted summation process on single-point scores of the plurality of target authentication points according to weights obtained from a testing of a trained training set; and authenticating the authenticity of the item according to the single-point scores or/and the total score. The problem that counterfeit goods are not be quickly authenticated can be solved effectively through identifying target authentication points used to describing the item in the images by convolutional neural networks, constructing an assessment mechanism, and determining the authenticity of the item through assessing identification results.