Digital curve identification from remote sensing data is difficult in continuous objects such as roads. Use of high spatial resolution images can increase geometrical details and accuracy of estimation and detect curvy segments from the road boundary. We detect a road as a curve in 2D raster grid and analyze its shape using fuzzy c-means and alpha shapes. Two approaches identify curvature from the polylines on two sides of the road. Image resolution, radius of alpha circles and size of moving window are the three main parameters for detection of curvy segments. Lower resolution, larger alpha circles and larger moving windows decrease the chance of detecting sharp and narrow curve segments.