In this paper, a procedure is proposed towards automated linear cartographic data acquisition from scanned topographic maps for GIS data entry. A bi-partition clustering algorithm is applied successfully to produce relatively clean and single featured map layers from the scanned topographic maps. Comparison with other clustering algorithms shows that our algorithm produces the best results and is the most efficient. In addition, cleaning and thinning are performed using the LINDA system in an automated manner, resulting in the final map layers, such as road, contour, drainage networks and railroad layers, ready to be input to GIS.