A Research For Ship Target Segmentation And Feature Extracting In High Spatial Resolution Remote Sensing Image

In remote sensing image, the segmentation of military targets ,feature extraction and the technology of automatic recognition are the hotspot and difficulty in the research of modern military information technology.They are applied mainly in many fields such as the collection of military information , monitoring and warning against the enemies and homing guidance weapons. In high-resolution(which involve spatial resolution, temporal resolution and spectrum resolution) satellite image, the image processing and recognition technology of ship targets have a great value in the military preparation against Taiwan and sovereignty problems in East Sea and South China Sea.They provide us with a significant information safeguard net on our oceanic sovereignty ,nautical interests and nautical battles against enemies.Meanwhile,in the modernization of our military ,they are urgency with the transition from mechanization to informationization.The satellite remote sensing images from the paper are the visible image taken by homemade"JBⅢ"military reconnaissance satellite.The spatial resolution arranges from 1 to 3 meters.This paper concentrates on the precision segmentation in ship targets of the remote sensing image and the feature extraction which comes from the ship targets after segmentation.Combining with the factual characteristics of the ship targets and background of the water,the paper firstly carry through a yawp removal and pretreatment of contrast enhancement,followed by an improvement on the current fuzzy enhancement algorithm in the processing of image enhancement. In the following process of image segmentation ,the paper focuses on image segmentation method which is based on threshold,then makes an segmentation experiment against ship targets by means of well-applied one-dimensional threshold segmentation method ,after that analyses the disadvantage.It spreads the feature space of the distribution pixels from one to two dimension by making use of gray-level information and spatial information in the process of segmentation.It indicates a good result by adopting the traditional two-dimentional Otsu threshold segmentation to segment the targets,while the shortcoming is that it has a complexing computation and is time-consuming. Finally,this paper presents a method that combined with region growing method , it segments images by means of a kind of simplified two-dimensional Otsu .The method divides the image segmentation into two processes .Firstly,it takes a initial segmentation by using simplified two-dimentional Otsu.Secondly,it carries on a precision segmentation by using region-growing method.Experimental result indicates that the method proves to be accurate ,rapid and well-applied.The paper is made up of five chapters , the keypoint of which are two ,three and four chapter.These three chapters make research respectively on the technology of image preprocessing ,image segmentation and feature extraction ,improvement and experiment which refered in this paper .