High-Speed Horizon Line Detection Algorithm Using Dual Hough Algorithm for Marine Image

Abstract

Horizon detection is the first step for analyzing video-based maritime situations on autonomous vessels. The detected horizontal line information is used for various aspects such as distance estimation, image registration, and the region of interest detection. Recently, a method of improving the accuracy of horizontal line detection by using multi-scale image and radon transform has been proposed, but there is a problem that detection time is long. In this paper, we propose a method to accurately detect the horizon at high speed for the marine image. We set a color-based region of interest, which is called ROI and uses Duale Hough Transform which improves the speed of computation and accuracy. In the experimental analysis, the proposed method shows similar accuracy for the horizontal line detection with MuSCoWERT and shows the similar processing time with Hough transform algorithm. The experimental results illustrate that both the processing time and accuracy can be improved together up to optimal with the combination of ROI and the Dual-Hough transform.