| Project Reference: | ITP/038/19LP |
| Project Title: | Deep-Learning AI Vision of Passenger Activity Patterns at the Airport to Enhance Travellers' Experience and Customs Facilitation |
| Hosting Institution: | LSCM R&D Centre (LSCM) |
| Abstract: | The Hong Kong International Airport is one of the busiest airports in the world and serving as the travellers’ hub for the region. With annually 75 million passengers passing through, the HK Customs steps up effort in safeguarding security of our airport, curbing smuggling activities to achieve a safe city. In the Customs statistics, dangerous-drug seizure has increased dramatically by more than 8 times in value, comparing the first quarter of 2018 and 2019. Smuggling cases are mostly by passengers carrying on in baggage. Customs carries out baggage clearance where only passengers of high risk are summoned for baggage check. Currently, passengers are observed for their patterns of activities, in which irregular patterns are considered risk factor. However, the manual procedures in it demand intensive personnel resources for the attention and investigation exercise. In order to enable speedy customs clearance service while effectively attending to suspicious cases, we propose to develop the technology to help automate the manual process. The proposal is to adopt the contemporary advanced results of video analytics on suspicious patterns of passengers’ activity around the baggage hall. Moreover, it is envisioned to enhance training of neural nets by mashing defined rules from Customs’ operational experience together with labelled video data. With novel implementation of these techniques, this project will demonstrate effective Customs operation at the same time facilitating smooth passenger flow in airport operation. Deliverables of the project will support the "Smart Customs" initiative of C&ED as well as contribute to HK as the efficient and safe port. |
| Project Coordinator: | Dr Frank Tong |
| Approved Funding Amount: | HK$19M |
| Project Period: | 01 Jan 2020 - 31 Dec 2021 |
