Obstacle detection for Unmanned Surface Vehicle

The goal of this paper is to present a new monocular vision system for obstacle detection to be used by USV (Unmanned Surface Vehicle). This system is based on a monocular camera placed on a known height above the waterline. The technique used to calculate the distance between camera and obstacle depends on earth geometry and requires the prior horizon line detection. The main contribution of this work is the improvement of the performance of horizon detection algorithm in term of calculation time by applying an initial transformation based on grayscale pre-processing. In fact, the idea came from observations made in maritime environment where the sky and sea colors are very close and the major edge and line detection techniques use grayscale images. Considering this case, we managed to significantly reduce calculation time under normal weather conditions (daylight vision and normal visibility). System experimentation was made with Matlab/Simulink for performance evaluation purpose. The proposed horizon detection algorithm was implemented with C++.