AUTONOMOUS SURVEILLANCE PLATFORM

Autonomous Surveillance Platform Capability Brief
Autonomous Surveillance Platform Capability Brief
Traditional surveillance methods rely on security personnel to observe and interpret surveillance data in real-time. Inevitably such methods do not scale well for busy environments, cannot merge data from disparate sources, introduce opportunity for human error and ultimately fail to provide adequate protection for modern transport corridors.

CSIT have developed a suite of technologies to deliver real-time situational awareness. Real-time analytic techniques are applied to video and RF data to detect and track human subjects through secure zones. Detections are combined with data acquired from traditional physical security infrastructure such as ticket verification and access control to detect and record events of significance and identify abnormal or inconsistent behaviours.

CSIT video analytics provide robust passenger profiling data such as age and gender. CSIT were the first to demonstrate real-time gender recognition based upon combined face and full-body data.

Spatial tracking is achieved by fusing multiple video sources with RFID to localise subjects and record movement trajectories. Appearance models and RF signatures for subjects are learnt in real-time to facilitate reacquisition as subjects move between secure zones.

ASP leverages the CSIT Evidential Reasoning Framework (ERF) to reason and fuse event data. A multi-agent architecture is adopted to devolve intensive analytics to dedicated agents and deliver fault tolerance and system scalability to meet the requirements for large deployments.

The ERF is commissioned with the necessary domain knowledge to detect abnormal event sequences reported by the agents. Where appropriate, the ERF can also access online data sources such as ticketing, POS or Access Control data to iden- tify observed inconsistencies.

The full capability brief is available to download from our website here - CB2-ASP (Autonomous Surveillance Platform) (pdf, 1.4MB)