Road Detection Based on Color and Geometry Characteristics

Many modern systems use road detection methods. The proposed automatic method for road detection consists of two steps. In the first step, the color segmentation in HSV color space is applied. The second step aims to improve the quality of color segmentation results by removing the false segmented areas. On the second step, the geometry features are used to classify segments into two classes - roads and background. We use four characteristics: shape index (SI), linear index (LI), elongation index (EI), eccentricity (Ecc). The geometry features are compared with the thresholds to classify segments. The presented results show the effectiveness of the proposed method.