Abstract: A new method for building extraction,without the need for DTM(digital terrain model) advance extraction,was proposed based on LIDAR(light detection and ranging) data and the shape analysis of contours.In this algorithm,firstly,the initial segmentation of DSM(digital surface model) contours is carried out based on the closure of a contour and the area enclosed by the contour,and the initial contours of buildings are acquired.These buildings are consisted of closed contours expressing building points and vegetable points and have a certain area.Secondly,after smoothing those initial contours,they are segmented to building contours and non-building contours based on dimension surveying and shape characteristics of the contours,such as rectangle tolerance and circle tolerance.Finally,the building contours predigested by the topological relation are served as the outlines of buildings and normalized based on the vertical or parallel characteristics of building outlines.In a test on an undulated region,buildings’ contours were extracted from DSM contours and regularized.To validate the test results,an overlay of extracted buildings’ outlines and an aerial image was checked.The result shows that the proposed method has an accuracy up to 80% and can effectively extract big and independent buildings in undulated regions.