<title>Automated anomaly detection processor</title>

Abstract

Robust exploitation of tracking and surveillance data will provide an early warning and cueing capability for military and civilian Law Enforcement Agency operations. This will improve dynamic tasking of limited resources and hence operational efficiency. The challenge is to rapidly identify threat activity within a huge background of noncombatant traffic. We discuss development of an Automated Anomaly Detection Processor (AADP) that exploits multi-INT, multisensor tracking and surveillance data to rapidly identify and characterize events and/or objects of military interest, without requiring operators to specify threat behaviors or templates. The AADP has successfully detected an anomaly in traffic patterns in Los Angeles, analyzed ship track data collected during a Fleet Battle Experiment to detect simulated mine laying behavior amongst maritime noncombatants, and is currently under development for surface vessel tracking within the Coast Guard's Vessel Traffic Service to support port security, ship inspection, and harbor traffic control missions, and to monitor medical surveillance databases for early alert of a bioterrorist attack. The AADP can also be integrated into combat simulations to enhance model fidelity of multi-sensor fusion effects in military operations.

Detection of vessel anomalies - a Bayesian network approach

“…Approaches for detection of anomalous vessel behaviour that do not require human expert knowledge of what is normal include the automated anomaly detection processor (based on a Self Organizing Map) described in [2], the trajectory clusterer reported in [3], and the fuzzy ARTMAP neural network classifier described in [1]. We agree that it is important not to rely completely on human expert knowledge due to the reasons mentioned above.…”

“…It has been argued that approaches for improving situation awareness should not require users to define templates or to specify prior knowledge of what constitutes normal/abnormal behaviour, since such knowledge often is not available and may change over time [2]. Approaches for detection of anomalous vessel behaviour that do not require human expert knowledge of what is normal include the automated anomaly detection processor (based on a Self Organizing Map) described in [2], the trajectory clusterer reported in [3], and the fuzzy ARTMAP neural network classifier described in [1].…”

Visual Analytics for the Detection of Anomalous Maritime Behavior

“…The majority of them present a fully automatic approach to anomaly detection problem, where the user is mainly considered as a consumer of information (e.g., the work presented in [16] and [2]). Examples of methodologies for anomaly detection that include human expert knowledge to any extent are rare.…”

“…The approach selected to build the normal/special behavior model is based on the work presented in [16] (a Gaussian Mixture Model (GMM) over a SOM of the training data is used for that). We have extended here their proposal, adding an interactive module that allows continuous refinement of the calculated model and development of a "special event" model by the user.…”

Vessel Pattern Knowledge Discovery from AIS Data: A Framework for Anomaly Detection and Route Prediction

“…In [17], a probabilistic model to track human behavior over time is presented. The papers [18][19][20][21] specifically deal with maritime applications, although using image processing techniques. Reference [12] presented an extensive model to statistically learn motion patterns without any prior knowledge in traffic scenes where the traffic flows are constrained to stay in specific areas.…”