Year
2001
Abstract
This paper describes our progress towards the development of an autonomous tracking system capable of monitoring the movement of transport vehicles and of detecting anomalous behavior. The system under development integrates commercial-off-the-shelf (COTS) GPS tracking hardware, a real-time database, a neural network algorithm for data analysis and anomaly detection, and a graphical user interface (GUI) for control and operation of the system. In order to track movement of transport vehicles from one location to another in real-time, we use both spatial and temporal features to input into the pattern recognition paradigm. The GPS system provides location and time data that can be used to establish an integrated network of virtual detectors. Each of these virtual detectors has a unique location and identifier, and is also a member of a \"local neighborhood\" of virtual detectors that are located in sequence along the possible paths of the transports. In many instances, it will be possible to identify anomalous movement over one of these local neighborhoods. This allows the real-time detection of potentially abnormal movement as soon as it occurs, and provides security forces with the opportunity to respond quickly and in an appropriate fashion. The system is designed to operate in one of two modes. The first mode is a training mode where the anomaly detection algorithm automatically learns normal transport patterns, in an unsupervised manner. The second mode is a tracking mode that will analyze realtime position data and monitor transport operations for anomalous activity. Both learning and tracking functions will be performed using a neural network paradigm. Once completed, this automated tracking system will be able to detect anomalous behavior and provide information on the nature of the anomaly, including the exact location of the transport. Incorporation of this tracking system into day-to-day transport operations at a site will reduce field support costs and will provide enhanced security.