Automatic Extraction of Road Network from Aerial Images

Abstract

Analysis of aerial images has been an important research topic for urban analysis. The classification and measurement for geometrical objects of a city will be simplified if its road network could be automatically extracted from aerial images. In this paper we present an automatic method to extract road network from high-resolution aerial images. These roads are found through the use of a consecutive linear features detection algorithm that is mainly based on the three following steps: first, pre-processing, a local selective window method followed by median filter are combined to remove noise and possibly irrelevant small scale details; which is followed by a cross entropy segmentation algorithm in variable overlapping window according to some rules, thus generating a binary edge map; in order to detect straight lines, a Hough transform is applied to the binary edge map to get the parameters of candidate lines. The experiment shows that the proposed method can be used to correctly extract road network from aerial images.