Turkish Journal of Electrical Engineering and Computer Sciences




Classification of marine targets using radar data products has become an important area for modern researchsociety. However, due to several reasons such as the similarity between ship structures and spatial specifications,classification of marine targets constitutes a challenging problem. In almost all of the studies, this problem has beenhandled by focusing on a single instance of range profiles or synthetic aperture radar data. However, this approachis seen to achieve only a particular success. This study introduces a novel classification approach that is shown toprovide additional classification enhancements by exploiting the extra information extracted from sequential rangeprofiles generated by ground-based marine surveillance radars. With this purpose, both synthetic and measurementalrange profiles are taken into consideration. Synthetic profile data are generated for seven marine targets by using anelectromagnetic scattering simulation tool (RASES)1. On the other hand, a total of 2387 range profile data of 171different target tracks are collected for five different marine target class types by using an X-band marine surveillanceradar. Each target tracked for a long period of time to gather sequential HRRP data subsets. HRRP data subsets areused to generate HMM based transition matrix probabilities and sequential classification results by evaluating proposedmethod. Probabilistic neural network (PNN) and convolutional neural network (CNN) classification algorithms appliedto gather classification results. The proposed method results are compared with both single value classification andmajority voting rule (MVR) method results. According to the examination results, the proposed classification approachprovides remarkable enhancements in the correct classification rates when compared to the case of single profile dataapproach.


Maritime automatic target recognition, sequential data classification, synthetic/measuremental rangeprofiles

First Page


Last Page