| Title |
A Study on the Complementary Use of Face Recognition Model and Re-identification Model for Efficient Identity Verification |
| Authors |
곽도환(Dohwan Kwak) ; 박윤하(Yunha Park) |
| DOI |
https://doi.org/10.5573/ieie.2025.62.8.72 |
| Keywords |
Face recognition; Re-identification; Deep learning; Biometric information; Surveillance system |
| Abstract |
This paper proposes a system to improve person identification performance by complementarily combining face recognition and re-identification (Re-ID) techniques. Face recognition is a technology that identifies individuals by analyzing facial biometric information, and it is widely used in various fields such as airport security, fraud prevention in finance, and surveillance systems. However, the accuracy of face recognition can degrade due to variations in lighting conditions, facial angles, and low-quality images. To address this issue, this study proposes a method that first extracts candidate groups using a full-body image-based Re-ID technique and then performs final person identification through face recognition. The face recognition model is based on the GhostFaceNet architecture and further improved by incorporating ResNet-based bottlenecks and SE modules. The Re-ID model uses a CLIP-based Re-ID framework, which integrates the attention results between text and images into the Re-ID process. Experimental results show that the combination of face recognition and Re-ID techniques significantly improves person identification performance in low-quality images and unconstrained environments, with clear improvements over baseline models. This study suggests that the integration of face recognition and Re-ID can enhance the practicality of various security and surveillance systems. |