Shaded Roads Detection Based on Contour Segmentation

Abstract

This paper deals with the detection of roads partially obscured by shadows that are cast by neighboring objects. The approach consists of three main steps: preprocessing, shadow detection and removal, and road detection. Detection of shadows and roads is based on contour segmentation in HSV color space. After contour segmentation, the H component is divided into segments. The value of a feature is calculated for each segment and is compared with the threshold. Brightness is used as a feature for shadow detection. Road detection is based on the color feature. The gamma correction method is used for shadow removal. Experimental results demonstrate the effectiveness of the proposed approach.