Application of generalized constrained energy minimization approach to urban road detection

Classification of thematic layers have been used to update digital databases, particularly digital maps which can be accessed on the Internet by a fast growing number of users. One of the most important features in digital maps is traffic patterns and road networks. Unfortunately, due to low spectral and spatial resolutions of satellite imagery, many roads cannot be extracted at pixel level since the width of most urban and rural roads is less than the pixel size. In addition, road spectral signatures vary from pixel by pixel due to natural background. In this paper, an approach combining a subpixel detection method, called generalized constrained energy minimization with a principal component analysis-based fusion technique is proposed for urban road extraction. Experimental results show that the proposed method can effectively detect roads in Landsat TM and SPOT panchromatic images