Road Extraction Of Remote Sensing Images Based On Support Vector Machine

With the wide application of remote sensing technology,the extraction of remote sensing image information has became a hot topic for many scholars.The computer image processing technology applied to the extraction of road information from remote sensing images,compared to traditional field mapping,it can greatly improve time efficiency,and with high application value for scientific research and civil use,it can greatly reduce the number of manual operations.At this stage,with the significant increase in the ability of satellites to acquire data,the resolution of acquired data has also been greatly improved,and we can distinguish buildings,roads and other related information from the remote sensing images captured by satellites.It is of great significance for the research and application to deal with the related information and obtain the region of interest.This paper mainly extracts road information from remote sensing images.In the remote sensing images acquired by satellites,we firstly use the improved GVFSnake segmentation algorithm to segment the road based on the road's inherent shape,and support vector machines(SVM)Training to classify road and non-road information based on the various characteristics of roads in remote sensing images and related information.Finally,a new contour skeleton extraction algorithm is proposed to extract the road skeleton and extract the road information.In order to segment the roads in remote sensing images,an improved algorithm based on GVF-Snake is proposed in this paper,which can segment the interested roads to facilitate the analysis of road features.Because the GVF-Snake algorithm is based on the image gradient,it is easy to distinguish the edge of the road when the edge of the image is blurred.Correspondingly,we propose to iteratively constrain GVF-Snake using the minimum width of the road.Experiments have proved that it can well limit the iterative problem of GVF-Snake at the edge of the road.In order to obtain road information,this paper divides the information in remote sensing image into road and non-road information.We use support vector machines for road and non-road classification operations.For the acquisition of the road section,this paper uses the improved GVF-Snake algorithm to segment and extract the road information,and manually select non-road information in the remote sensing image,and we use the two types of obtained information as training data.By using the two types of training data,road information and non-road information to train the support vector machine,we can complete the classification operation of roads and non-roads in the satellite remote sensing images,and achieve the extraction of road information.This paper proposes a new contour skeleton extraction algorithm.In this paper,we first study the traditional skeleton extraction algorithm,which is the largest disk algorithm in OPENCV,simply extracts the skeleton from the image and does not consider the connectivity of the skeleton.We first use the Hough transform to detect the extracted skeletons,determine which points in the image are located in the same segment by straight line detection,and judge the line segments,the line segments indeed consistent with the connectivity.We use extensions to connect those disconnected due to the blurry image information,and we use the expansion algorithm to connect the local area,which is a part of the curve.And eliminate glitches.Finally,we use the above-mentioned three steps to perform road extraction on remote sensing images in order.The extraction results can basically satisfy the use,which proves that the improved algorithm proposed in this paper improve the road extraction efficiency and has a higher accuracy.The realization of the algorithms in this paper has a certain significance in the field of scientific research and application.