Application of a Fast Linear Feature Detector to Road Extraction From Remotely Sensed Imagery

Abstract

This paper presents a fast and effective algorithm for detecting ridge- or ribbon-like linear features from remote sensing imagery. To judge if a pixel is at the center of a linear feature, the first step is to find several biggest pixels by their grey values within orthogonal directional windows, and store them into an evaluation window. A simple evaluation method is then applied to make the yes or no decision on whether the pixel is a linear feature point. Aerial images were used to test the algorithm's ability to extract roads. This algorithm was compared with the Multiple Directional Non Maximum Suppression (MDNMS) algorithm. The experimental results indicate that with the proposed algorithm the processing of road details could improve and the processing time decrease.