| Project Reference: | ITP/034/22LP |
| Project Title: | Video Analytics for Tracking Cyclists in Track Cycling Race from Dynamic Pan-Tilt-Zoom Scenes |
| Hosting Institution: | LSCM R&D Centre (LSCM) |
| Abstract: | This seed project aims to digitalize track cycling race data from pan-tilt-zoom filmed videos for paving the way to deepen data-driven sports analytics for strengthening the support for Hong Kong Cycling Team to derive training and game strategies. The R&D methodology involves the integral use of techniques in machine learning, computer vision, and information science. A cyclist detector viewed from multiple angles will be trained through a deep learning mechanism. Relevant computer vision algorithms will be used to track detected cyclists across the pan-tilt-zoom (PTZ) scenes embedding with moving background and encountering multiple viewing planes. Information science techniques will be used to keep track of the spatiotemporal relationships between each of the local PTZ scenes to a re-constructed global race track. Together, cyclists’ spatiotemporal positions to the race track can be obtained for analyzing their interactions during a game, which is the insight for strategizing game plans. The R&D work on tracking moving object with a moving PTZ camera, instead of a static camera, may lead to more opportunities to apply video analytics over visual content from drones or other PTZ applications for transportation, manufacturing, logistics, surveillance, as well as sports industries. |
| Project Coordinator: | Dr To Bun NG |
| Approved Funding Amount: | HK$ 2.79 M |
| Project Period: | 1 Dec 2022 - 29 Feb 2024 |
