Railway contactless checkout process with identification assisted by gait recognition

Area of Science:

• Computer Science
• Biometrics
• Artificial Intelligence

Background:

• Railway stations are optimizing passenger entry and exit procedures for efficiency and convenience.
• Current identification methods include manual control, ID cards, magnetic tickets, and emerging face recognition technology.
• There is a need for improved contactless identification to further enhance passenger throughput and security.

Purpose of the Study:

• To propose an advanced contactless checkout process using gait-augmented face recognition.
• To develop and validate a weakly-supervised body segmentation network (Dwsegnet) and an improved GaitSet model.
• To enhance the accuracy and efficiency of contactless identification in railway passenger services.

Main Methods:

• Development of Dwsegnet, a weakly-supervised body segmentation network.
• Implementation of an improved GaitSet model for gait recognition.
• Integration of gait and face recognition for a multimodal identification system.
• Comparative analysis against existing identification models.

Main Results:

• The effectiveness of Dwsegnet and the improved GaitSet model was validated through comparative studies.
• Gait-augmented face recognition achieved a 2.31% improvement in contactless identification rate compared to single-modal face recognition.
• The proposed method demonstrates superior performance in contactless identification accuracy.

Conclusions:

• Gait-augmented face recognition offers a significant improvement over single-modal face recognition for railway passenger checkouts.
• The developed Dwsegnet and improved GaitSet models are effective for enhancing contactless identification.
• This innovative approach contributes to more efficient, secure, and convenient passenger processing in railway systems.