Quality of target tracking in marine radar has a great impact on the safety of navigation. One of the most important challenges in tracking is to estimate movement parameters for maneuvering target. Various algorithms for this purpose have been proposed, including multiple model approach. One of the alternative for commonly used numerical filters may be the use of artificial neural networks. Earlier research has shown that it may be useful to track targets with neural filters. In the research project called "Development of marine target tracking methods with the use of neural filtration" multiple model neural filter, based on General Regression Neural Networks has been proposed. Such a filter was implemented in dedicated software for testing. One of the steps of verification of the method is verification in simulation environment. This research are presented in the paper. Two kinds of simulators have been used and the results was tested against IMO requirements and compared to commercially used radar. The results have confirmed that proposed method complies with IMO requirements, and it may be also competitive to commercial solutions. The research was financed by polish National Science Centre.
展开▼
