Research On Road Extraction And Representation From High Resolution Remote Sensing Images

Road is important basic geographic information in the field of traffic and transportation. In recent years, with the development of high resolution remote sensing satellite launch, how to extract the road from high resolution remote sensing image is a very challenging task. Therefore, it has become one of the hotspot in some fields, such as remote sensing, computer vision and image understanding.This paper focused on the road extraction method based on high resolution remote sensing image. The main research contents of this thesis are as follows:(1) An automatic road extraction method of remote sensing image based on color line segment detector is proposed. First, gradient of HSV color model is used in line segment detector. Then, a set of line combination rule and automatic method for selecting line and candidate road region are designed.The experimental results show that the proposed algorithm has a better road extraction performance compared with previous methods.(2) A semi-automatic road extraction algorithm of remote sensing image based on the constrained region growing is proposed. The constrained region growing algorithm is presented by imposing three constraints on the classic region growing method: 1) for the growing area, the growing region is chosen from road class which is obtained by K-mean clustering; 2) for the growing direction and rules, the single growing direction is replaced by multi-direction growing, and nine road intersection models is presented to constrain the growth; 3) for the threshold and decision rule, the threshold is set as the standard deviation of the roads class, the decision rule is updated as comparing the distance between each road class pixel and the centroids with the threshold. Experimental results validate the feasibility and effectiveness of the proposed method.(3) After the road network has been extracted, the distance between pixels is calculated as length and width of the straight line road(surface), and parabola and numerical integra1 is utilized to calculate the length of curved road. Meanwhile, the application of the extracted road information isdiscussed. Effectiveness of the road information expression is shown in the application.