Improving the Maritime Traffic Situation Assessment for a Single Target in a Multisensor Environment

Abstract

The key to maritime surveillance is an accurate and real-time update of the current traffic situation. Although the Automatic Identification System (AIS) has greatly improved the traffic situation assessment over the past years, it has shown to be error-prone and vulnerable to spoofing or intentional misuse. To obtain a more reliable picture of a vessel’s motion this work proposes the fusion of AIS and Radar data in a loosely coupled architecture. For that reason, an Interacting Multiple Model (IMM) Multi Sensor Probabilistic Data Association (PDA) filter was designed not only to fuse measurements from different sources, but also to best capture the different dynamic states of a single vessel. In the IMM two Unscented Kalman Filters (UKFs) run in parallel and interact with each other, each implementing a different dynamic model, the CV and CTRV, respectively. Each UKF is conditioned on asynchronous measurements obtained from either AIS or Radar. In case of the latter, measurements of a target candidate may also originate from clutter, not from the target itself. The PDA filtering approach not only associates these measurements to the actual target state but also acts as innovation gate for faulty AIS position updates. A standard blob detection algorithm was implemented and tuned for extracting target candidates in range and bearing from Radar images. To show the benefits of the proposed scheme a dedicated measurement campaign was performed in the Baltic Sea with a stationary ship monitoring a highly manoeuvring target vessel. With the proposed approach not only short term AIS induced errors can be detected and rejected but what may also become observable is suspicious AIS behaviour including anomalous position data or muted transponders.