Automatic road detection using MCSC

Roads are an important part of life for travelling & transportation system. Map is the tool which is used commonly for navigating and recognizing the roads. Due to sprawl of population, changes happen within or in suburbs of cities. We have to redesign the map(s) for such situation and manually updating the map is very complex and time consuming task. For automatic map generation from satellite images it's essential to extract the roads first We propose a new method Multiple Simple Color Space Components (MCSC) to detect the road region(s) from satellite images. In our research we experimented with multiple color space components to extract the roads of Mirpur city by using satellite images (SI) from Google Earth. We used selective components of various color space models namely YCbCr, HSV and L*a*b*. To obtain the road region(s) from satellite images different steps were followed i.e. feature extraction, segmentation and grouping. For extracting feature, we used following color components (i.e. Luminance (Y), Saturation (S), Hue (H) and chromaticity layers `a*' and `b*') on different satellite images. Segmentation was done by Thresholding and multiplication of the S, H, a* and b* images with each other to eliminate the non road regions. Result of this process (H-S and a*-b* images) are combined with the luminance (Y) to detect the road region. The proposed MCSC processing method can detect roads easily and generate results quickly within a second. It's a very simple, fast and fully automatic algorithm to detect the road(s) region(s). Furthermore, the proposed system also gives good results in complex environments/backgrounds.