Research Of Road Extraction From High-resolution Remote Sensing Image

With the development of remote sensing technology,more and more information is contained in remote sensing images.Compared with traditional field mapping,using computer technology to acquire target information from rich remote sensing images has the advantage of high efficiency and low cost.Road is the main component of social basic geographic information,and is one of the fastest urban renewal information.The rapid acquisition of road information has important political,military and economic significance.In this paper,a large number of documents are consulted,and the research status of road extraction at home and abroad is analyzed and summarized.On the basis of fully affirming the existing research,the shortcomings of various road extraction methods are pointed out,and the research on the automatic extraction of roads from high resolution remote sensing images is carried out in view of the shortage of current road extraction methods.The research scheme of this paper is mainly divided into image preprocessing stage,low level road information extraction stage,middle level road feature analysis and image feature set construction stage,and high level support vector machine classification stage.First,in order to reduce the influence of other objects in remote sensing images on road targets,the images are preprocessed in this paper,including remote sensing image correction,image fusion and image enhancement.Then,the characteristics of the road target in the image are analyzed in the middle level processing stage,and how the Delaunay triangulation algorithm distinguishes between the road class and the non-road class is described.From the three aspects of the radiation characteristics: geometric features and texture features of the road,the image feature sets of several triangles(defined as the triangle elements of the image)are constructed,and six characteristic parameters are described in red band,green band,blue light band,road width,triangle element border length ratio and trigonometric element two order moment.In the high level processing stage,this paper classifies all trigonometric elements in the Delaunay triangulation network by using the classification method based on support vector machines,and extracts the path target information.In order to verify the reliability of this method,several road extraction experiments are carried out on remote sensing images and UAV images.The experimental results show that the algorithm not only extracts the results quickly and accurately,but also extracts road information according to the material and the width of the road.