The reconstruction of buildings provides useful input for tridimensional city models generation which represent the basic layer of the Smart Cities concept for implementing high technology in the urban environment. ALS point clouds are mainly used as data source for mapping in different application fields. Before a building can be mapped it is necessary to be detected in the point cloud through an automatic procedure. The development of detection methods is a complex task due to significant diversity of the elements, the random structure of point clouds and the characteristics variability of point clouds created by the airborne laser scanner. For the study area the academic campus was chosen being a complex area which is composed by several classes: buildings, roads, high, medium and low vegetation, water course. The point cloud contains predefined classes created based on signal analysis, but because to a low confidence of this attribute, the surface reconstruction has to be done taking into consideration the geometry. The outliers together with the noise were removed after point cloud filtering and additional attributes have been calculated for each ALS point. This procedure must be carried out before the point cloud segmentation, for which region growing was chosen, being followed by the extraction of roof edges through least square line fitting. Automating the process of generating 3D objects improves the methodology of extracting geometrical 3D information with a high precision and accuracy.