Automatic Road Digitizing of Segmented Aerial Images for Urban Areas Based on K-means and Hough Transformation

Abstract

In Geographical Information System (GIS), digitizing is an exhausting and time consuming process which is used for digital mapping and map updating. Popularity of mobile services and autonomous vehicles technology based on roads network increase the need for road automatic digitizing. Last years, deep-learning gave excellent results on roads segmentation from aerial and satellite imagery [1], that highlighted the automatic digitizing. In this paper statistical method is applied to extract roads from segmented images, and tested with images for three different scenes and gave an average accuracy of 86 percent.