Object Detection and Tracking System with Improved DBSCAN Clustering using Radar on Unmanned Surface Vehicle

Abstract

Unmanned Surface Vehicle(USV) is a promising solution for missions that happened on the ocean such as patrolling, rescuing. It is necessary to detect obstacles for autonomous navigation. However, marine radar has some limitations, which is normally used for USV. Low update rates and the dead band for the local area concerning sensors are weak points, so that USV can hardly cope with close obstacles with high speed. Compares to marine radar, FMCW radar has opposite features. Hight update rates and available for close obstacle detection. This paper proposes an algorithm for detecting close obstacles with FMCW radar using the clustering method. Suggested algorithm compute spatial data from the Range-Doppler map given by FMCW radar. And apply Density-based spatial clustering of applications with noise(DBSCAN) algorithm considering radar energy signal level. Using radar data, the algorithm computes representative clusters and track the clusters with the Nearest Neighbor algorithm. This paper also shows the field experiment result of the proposed algorithm with USV. The field experiment is conducted in Chungcheongnam-do Taean Coastal Pier with USV called Sea Sword III. The results show obstacles are well detected and USV behaves collision avoidance.