Study On The Method Of High Resolution Remote Sensing Image Road Extraction

As more and more access to remote sensing image,and the image resolution is higher and higher,we obtain the target image is becoming more and more easy.But at present,the extraction of image information is mainly depends on artificial interpretation,this method is inefficiency,and the accuracy of interpretation has a lot to do with the professional ability of the interpreter.Therefore,how to realize fast and exact semi-automatic or full automatic remote sensing information extraction has great significance.This paper mainly studies the semi-automatic extraction of roads in the high resolution remote sensing image,road information is an important basic geographic information,is also one of the fastest update basic geographic information in city.The updating timely of the road has a very important role for urban planning and construction,traffic management,GIS data acquisition,spatial database update,image understanding.so the realization of rapid and accurate extraction of roads has very important significance.Before the road of image are extracted,we must first have a series of image preprocessing,mainly includes image segmentation and median filtering processing.This paper first adopts different segmentation method for image segmentation,select the most appropriate method for this experiment,then use the median filter to filter the image which has been segmented to remove the noise points.During the image preprocessing,according to the characteristics of the image and extract conditions of the road,also contain linear stretching,remove small area objects connected domain processing,etc.This paper mainly research on the accuracy of mathematical morphology and the Hough transform these two methods in extracting the road in the remote sensing image.Mathematical morphology is the mathematical methods for analyzing the geometric shape and structure,it is a kind of nonlinear operator,composed by a set of the morphology's algebraic operator command.It has four basic operation:inflation(expand),corrosion(erosion),opening and closing.Mathematical morphology algorithm is simple and flexible,fast speed,easy operation,and this method can optimize the image on the basis of holding original information,highlighting the image geometric features in order to further analysis.Classical Hough transform is a method used for detecting the graphics in the binary image,such as detecting straight line,round,oval,etc.The most classic application of Hough is the straight line detection.Hough transform have good adaptability for image noise,edge character of discontinuous and topography deformation.The algorithm is stable,effective and high reliability.Therefore this paper respectively uses the mathematical morphology and the Hough transform those two methods to extract roads in high resolution remote sensing testing image,then analyze the advantages and disadvantages of those two methods,and assess the extraction accuracy.Finally,combined with mathematical morphology and the Hough transform two methods,take the Resources satellite three image's panchromatic band as experiment object,extracting the road in it,and evaluate the road's extraction accuracy.