A Coarse-to-Fine Horizon Detection Method Based on Between-Class Variance
Yang Gui, Dongsheng Lin, Shaojun Liang, Luoting Xu, Junxin Hong · 2016
A coarse-to-fine horizon detection method based on between-class variance is present. Firstly, the original image is resized to a smaller one and plenty of straight lines are selected acoording to two line parameters which are distance to image center and inclination angle in the resized image. A criterion based on the between-class variance is set up for determining the best straight line which is considered as the coarse horizon. Then several horizontal coordinates are selected in the original image equally. The points corresponding to the horizontal coordinates which are the most likely on the real horizon are found nearby the coarse horizon according to the edge information and between-class variance. Finally, RANSAC algorithm is applied to preserve interior points which are really on the horizon and eliminate exterior points which are not, and line parameters of the real horizon can be gained by applying least squares line fitting algorithm for all interior points. The experimental results show that the proposed method can detect out the horizons under various complicated backgrounds effectively and has many advantages such as speedy calculation, strong robustness and high engineering value.