AI-based traffic anomaly detection
- Project
- 18036 SMART
- Type
- New system
- Description
The component for AI-based traffic anomaly detection detects a wide range of traffic anomalies in real-time from a given surveillance camera video sequence. The following traffic anomalies can be detected by the trajectory-based detector: zig-zag driving, driving in opposite direction, improper lane usage, road parking, side parking improper crossing by pedestrians or bicyclists, etc. The video-based anomalies include accidents, littering, fighting, robbery, etc.
- Contact
- Egor Bondarev
- e.bondarev@tue.nl
- Research area(s)
- AI, computer vision
- Technical features
The system runs in real-time mode and analysis the video feeds on the fly. It provides anomaly notification messages that include: anomaly type, location, timestamp, and video segment.
- Integration constraints
The system requires GPU of a 2080Ti type or higher, 12GB of memory.
- Targeted customer(s)
Providers of security/surveillance solutions, TMS integrators, camera manufacturers
- Conditions for reuse
Licensing
- Confidentiality
- Public
- Publication date
- 16-10-2023
- Involved partners
- ViNotion BV (NLD)
- Eindhoven University of Technology (NLD)