Research On Method Of Docked Ship Detection Based On Visual Saliency

With the application of new sensors and the development of space technology,the resolution of remote sensing image is getting higher,and these high-resolution images has widened the application field of remote sensing image processing,and put forward great challenge.Port is one of the important aspects of remote sensing image application research.It is of great practical and urgent significance to automatically extract the port ship target in the remote sensing image,which is important for shipping service and military intelligence.As a hub of the ships' target area,port contains a large number of target information,and as the waterway border area,the background is more complex than the sea.Due to the complex background and the polymorphism of port,the port ship detection is highly susuceptible to the background environment.In this paper,matching method is used to get the ROI area in the complex background,and then the visual saliency is used to extract the ship targets.The contents are as follows:1)This paper presents an improved visual saliency method based on global contrast algorithm.This method repositioning the image center,and introduces a new spatial weighting factor according to the characteristic that the obvious target area is more attractive than the surrounding.The evaluated result show that the improved algorithm has a better performance when the target is in the edge area of the image.2)This paper proposes a method to locate the port based on a priori location information.Aiming at the problem that ship detection is susceptiable to the complex background,the target ROI region is extracted by SURF feature matching with the priori location information.Experiments show that this method can effectively realize the separation of land and sea,eliminate the land interference,and make the basis for ship detection part.3)A method of ship detection based on the improved visual saliency is proposed for the ROI image area.The improved visual saliency is used to enhance the target region,and then the target contact line is determined according to the “peak-trough-peak” structure of the adjacent ship pixels.This method can deal with the shadow and the multi-target accumulation to achieve effective segmention of the targets.