Research On Key Technology Of Road Extraction And Change Detection For High Resolution Remotely Sensed Imagery

With the increasing use of high resolution remote sensing images in environmen-tal,urban,ecological,agriculture applications,the acquisition of more comprehensive,richer and more accurate information from remote sensing images has become the main trendency in remote sensing image analysis.Road information plays an important role in national life,transportation industry,and geographic information science research.It makes road information extraction the key issue in remote sensing applications.How-ever,road information is constantly in a condition of changing and updating,which makes the manually extraction a great deal and ineffective of work.Therefore,it is nec-essary to develop new techniques for road information extraction from high-resolution remote sensing image by exploring remote sensing technology,electronic technology and image recognition technology,and make it more effective and efficient.To meet the requirements for road information extraction application,this disser-tation studies the road information extraction of remote sensing images,form theory,method and applications respectively.The main works are listed as follows:1)A road image pre-processing method is proposed,which involves the following content:(1)Based on the view of road extraction from high resolution remote sensing im-age,taking into account the noise of image in pre-processed stage,this paper introduces an image enhancement method based on the joint image registration and point spread function(PSF)estimation super resolution(SR)reconstruction.Though the multi-frame SR,data source for sophisticated geometric feature and rich texture feature of road is provided.Firstly,the joint SR approach is formulated as a convex optimization problem which minimizes the combination of geometric motion and radiation difference;and then PSF estimation and image registration are achieved simultaneously and pro-gressively,to handle the error in different levels;finally,an iterative scheme based on alternating minimization(AM)is developed to solve the presented cost function.The experimental results confirm the effectiveness of the proposed method.(2)To overcome the shortcoming of traditional methods in large-size remote sensing image registration and fusion,a rapid image registration of remote sensing image is acvieved by using MPICH parallel model,and the a improved Pansharp method of remote sensing image fusion is proposed.The experiment results show that the proposed algorithm performs well on both efficiency and accuracy.2)An adaptive multifeature sparsity-based model for semiautomatic road extrac-tion is proposed.This dissertation studies and introduces the extracion of road central point by kernel mean shift tracking with the initial road position.Specially,the re-search is conducted from the perspective of feature representation and feature extrac-tion:firstly,a multi-feature sparse model is introduced to represent the road target appearancethe,which aims to maximally maintain the basic discrimination between the road target and the background;secondly,a robust and effective sparse constraint regularized mean-shift algorithm is used to support the road tracking.The experiments confirm that the proposed method performs better than the current state-of-the-art methods for the extraction of roads from HR imagery,in terms of reliability,robust-ness,and accuracy.3)With the help of vector data,this paper proposes a method for the roads changes detection.Firstly,road information is obtatined using the conbination of edge extraction method,template matching method and target tracking method;and then road change information is obtained by analysis and comparison with vector data;finally,road change information on the multi-temporal images is used to obtain the road change map.Multi-temporal high resolution images are used as experimental data,and the effective change detection results are obtained.