Research on Seaskyline Detection in Complex Sea Background

In this paper, several sea-sky-line detection methods, such as Hough transformation, straight line fitting and improved Otsu method, are compared. To improve the shortages of them, a new combined algorithm is presented based on the linear model of sea-sky-line, which is divided into four steps: 1) Background-suppressed by median filtering and gray extending. 2) Pre-segmented by selecting gray threshold based on the maximum variance between classes. 3) Using edge detection to keep the points in the obvious gray changing area. 4) Detecting the sea-sky-line by the Hough transformation. In step four, we also compare the results which are obtained by Hough transformation with those by straight line fitting. The experiment results show that our method can detect the sea-sky-line exactly and effectively under some complex background both in the condition of visible light and infrared ray. Besides, it also has strong robustness and higher detection efficiency to single frame.