Data Analytics, Machine Learning and Risk Assessment for Surveillance and Situation Awareness
Modern surveillance networks are able to provide trajectories of all kinds for aircrafts and vessels worldwide or at least in extended areas of the airspace or earth surface. Best known are Automatic Dependent Surveillance – Broadcast (ADS-B) and (Satellite-) Automatic Identification System (AIS) used in air and maritime surveillance. Both of them are cooperative systems. Besides these sources, sensors based on ground installations or mounted on airborne and space-based platforms deliver object trajectories independently of any transponders. This is done by advanced tracking and fusion algorithms generating trajectories out of sensor measurements. Examples include GMTI radar-based systems operating on UAV platforms or imaging systems based on high altitude pseudo satellites (HAPS) and satellites. These surveillance systems enable a continuous extraction of mid- and long-term trajectories of objects. Besides the trajectory generation, the challenge will be to place them into the right context and to provide situational awareness. This includes the estimation of the intents of the tracked objects, activity-based intelligence, and the determination of patterns of life. Otherwise, even modern surveillance systems are not able to take a real advantage of the gathered data. Therefore, trajectories are further processed by data analytics and machine learning. Unsupervised machine learning offers techniques to cluster and to partition trajectories, extract highly frequented routes and points of interest, predict object movement and identify anomalous behaviour. On the other hand, transponder and broadcast systems provide additional attributes of the tracked trajectories. These labels pave the way for numerous supervised machine learning methods. The derived predictors realise the determination of object types and activities. Finally, these new data analytic techniques have to be integrated in existing near real time surveillance systems. This requires specific system architectures as well as a completely new software and hardware landscape. In summary, trajectory-based data analytics, machine learning and risk assessment are embedded on local or global clouds and use dedicated mechanisms for distributed and parallel processing
Paper
Full text
Data Analytics, Machine Learning and Risk Assessment for Surveillance and Situation Awareness
Semantic Scholar · Computer Science · 2020
Abstract
Modern surveillance networks are able to provide trajectories of all kinds for aircrafts and vessels worldwide or at least in extended areas of the airspace or earth surface. Best known are Automatic Dependent Surveillance – Broadcast (ADS-B) and (Satellite-) Automatic Identification System (AIS) used in air and maritime surveillance. Both of them are cooperative systems. Besides these sources, sensors based on ground installations or mounted on airborne and space-based platforms deliver object trajectories independently of any transponders. This is done by advanced tracking and fusion algorithms generating trajectories out of sensor measurements. Examples include GMTI radar-based systems operating on UAV platforms or imaging systems based on high altitude pseudo satellites (HAPS) and satellites. These surveillance systems enable a continuous extraction of mid- and long-term trajectories of objects. Besides the trajectory generation, the challenge will be to place them into the right context and to provide situational awareness. This includes the estimation of the intents of the tracked objects, activity-based intelligence, and the determination of patterns of life. Otherwise, even modern surveillance systems are not able to take a real advantage of the gathered data. Therefore, trajectories are further processed by data analytics and machine learning. Unsupervised machine learning offers techniques to cluster and to partition trajectories, extract highly frequented routes and points of interest, predict object movement and identify anomalous behaviour. On the other hand, transponder and broadcast systems provide additional attributes of the tracked trajectories. These labels pave the way for numerous supervised machine learning methods. The derived predictors realise the determination of object types and activities. Finally, these new data analytic techniques have to be integrated in existing near real time surveillance systems. This requires specific system architectures as well as a completely new software and hardware landscape. In summary, trajectory-based data analytics, machine learning and risk assessment are embedded on local or global clouds and use dedicated mechanisms for distributed and parallel processing