Sea-sky line detection method based on scene division

A detection method of sea-sky-line based on scene division was proposed, taking into account two types of scenes and improves the accuracy while avoids the unnecessary time consumption. Firstly, a convolution operation was used to divide the preprocessed image into a simple scene or a complex scene. An appropriate method for detection was selected according to the result of the division. For simple images, first using a local Otsu algorithm based on image segmentation, and the abrupt pixels were extracted as edges. The sea-sky line was detected by using a Hough transform. The significant targets were eliminated after segmentation for the complex images. The morphological filtering operation was used to remove the interfering pixels and detect the sea-sky line. The experiments show that the proposed method can effectively improve the accuracy of sea-sky line detection and has good scene adaptability. © 2020, Editorial Board of Journal of Huazhong University of Science and Technology. All right reserved.