On Background Subtraction for Target Tracking in Complex Multi-Radar Marine Data Set

Abstract

Maritime surveillance and traffic control heavily rely on Automatic Identification System (AIS), which can be easily manipulated and is dependent on GNSS, therefore they are susceptible to jamming and spoofing, especially in areas of high economical and geopolitical interest, e.g. the Baltic Sea. The development of frameworks to augment the traffic assessment, such as automated Multiple Target Tracking (MTT) algorithms on maritime radar, is currently dependent on synthetic data where common scenarios present in real-world data are not properly emulated. This work presents a collection of data sets from X-band radar ground station, and their respective AIS data, installed in a busy port region with several challenging scenarios for data processing and target tracking providing real- world cases for MTT development and evaluation. The present data set is heavily contaminated by static structures with no relevance for surveillance and traffic assessment, which can be considered as a background layer. Thus, two methods are presented and evaluated to filter the radar pixels related to background structures in order to improve MTT algorithms.