Font Size: a A A

The Study Of Ship Model Recognition In Optical Remote Sensing Images

Posted on:2017-06-05Degree:MasterType:Thesis
Country:ChinaCandidate:S X ZhangFull Text:PDF
GTID:2392330590969409Subject:Aeronautical and Astronautical Science and Technology
Abstract/Summary:PDF Full Text Request
Our country's maritime territory is vast,using optical remote sensing images to monitor the oceans has great significance.This paper studies the ship target model identification,including sample library construction,image preprocessing,feature extraction and description,research for recognizing algorithm.1)Construction of ship sample library.First of all,this paper collects the basic information of five different classes of ship-targets.Then collects nearly 300 ship sample from the Google Earth,a complete sample library of ships is constituted fianaly.2)The preprocessing of ship target.The Ostu segmentation is fist performed,then to solve the problem of irregular direction,the main component analysis algorithm is proposed to extract the spindle direction of the ship.And for more complete contour of ships,scan line filling algorithm is proposed to fill the segmented ship images.Finally,ships that parked side by side are split by looking projection histogram valley point.3)The classification of ship target.Firstly,ship's size,shape and texture features are extracted respectively,and generate different combinations of features.Then through the support vector machine classifier by using these features and features' combination,the classification experiments were carried out.Then analyze the classification results to get the best combination of features for ship classification.4)The recognition of ship model.The destroyers are close to each other in size,and similar in shape,so are frigates.To recognition them,a ship recognition method based on feature matching is proposed,including SIFT(Scale Invariant Feature Transform)and Harris corner.First,the local features of a ship are extracted by SIFT operator.To reduce the mismatching of SIFT,a ship is divided into three regions(the bow,the ship amidships and the aft)and then respectively match corresponding region.Secondly,extract the features of the Harris corner,which describe the edge of the ship,and then perform the matching.Later different weights according to the importance of the two matches were given,and finally determine the type of a given ship.
Keywords/Search Tags:Optical remote sensing image, Ship target, Model recognition, SIFT features, Harris Corner
PDF Full Text Request
Related items