Related Experiment Video
Updated: Jan 11, 2026
09:50
Author Spotlight: Eco-friendly Photoluminescent Textile Authentication with Curcumin
Published on: December 22, 2023
2.2K
Recognition model for counterfeit protection system in colour-laser-printed documents based on improved ShuffleNet
Qi-Ming Zhou1,2, Lu-Wen Yuan3, Qian Zhou2
1School of National Security, People"s Public Security University of China, Beijing, 100038, China.
Scientific Reports
|November 19, 2025
Summary
This study introduces an improved ShuffleNet_OD_CA model for enhanced Counterfeit Protection System (CPS) recognition in color laser printers. The model achieves 91.18% accuracy, outperforming traditional methods with fewer parameters.
Area of Science:
- Computer Vision
- Forensic Science
- Machine Learning
Background:
- Counterfeit Protection System (CPS) examination is vital for identifying color laser printers.
- Current manual inspection methods for CPS recognition are inefficient and inaccurate.
Purpose of the Study:
- To develop an improved ShuffleNet_OD_CA model for enhanced CPS recognition in color laser-printed documents.
- To improve accuracy and efficiency in identifying printer brands and individual printers.
Main Methods:
- Developed an improved ShuffleNet_OD_CA model based on the ShuffleNet V2 lightweight convolutional neural network.
- Constructed a dedicated dataset using documents from eight major color laser printer brands.
- Trained and tested the model on the constructed dataset.
Main Results:
- The enhanced ShuffleNet_OD_CA model achieved a recognition accuracy of 91.18% on the test set.
- The model has 1.82 million parameters and 80.3 million FLOPs, fewer than the baseline model.
- Demonstrated superior recognition accuracy and fewer parameters compared to classical models like ResNet.
Conclusions:
- The improved ShuffleNet_OD_CA model effectively identifies CPS patterns in color laser-printed documents.
- The model offers a more accurate and efficient solution for CPS recognition compared to existing methods.
- The model's capability to extract and analyze key image features of CPS was demonstrated.
