In this project MICC is involved for developing the visual surveillance and camera management system. We developed three core algorithms to make the system deployable in the wild.
- Camera stabilization: to allow the system to work even on high-pole fixed cameras subject to wind and possibly hit by external objects;
- Object detection and tracking system: localizing every moving object in the parking area, classifying the object as a vehicle, person or other and obtaining a precise real-world trajectory;
- Anomaly detection module: exploiting local trajectories and feature to understand if a statistically unlikely pattern is happening in the scene.