A Siamese Neural Network for Non-Invasive Baggage Re-Identification

Area of Science:

• Computer Vision
• Artificial Intelligence
• Transportation Systems

Background:

• Baggage handling systems (BHS) face challenges with traffic jams caused by incorrect check-in entries.
• Efficient baggage re-identification is crucial for safer and faster airport operations.

Purpose of the Study:

• To develop a Siamese Neural Network model for accurate baggage similarity estimation.
• To enhance the re-identification capabilities within airport Baggage Handling Systems (BHS).

Main Methods:

• A Siamese Neural Network architecture was employed to learn discriminative features for image similarity.
• The model was trained on a publicly available suitcase dataset, allowing for varied image conditions.
• The model's performance was evaluated against state-of-the-art architectures.

Main Results:

• The proposed Siamese Neural Network model achieved high accuracy in estimating baggage similarity.
• The model demonstrated superior performance compared to the leading state-of-the-art architecture on the benchmark dataset.
• The network effectively learns features for robust re-identification across different image conditions.

Conclusions:

• The developed Siamese Neural Network model offers a promising solution for improving baggage re-identification in BHS.
• This approach enhances operational efficiency and safety in airport baggage handling processes.
• The model's adaptability to different pre-trained backbones suggests broad applicability.