|M.Sc Student||Abo Akel Nizar|
|Subject||Automatic Road and Terrain Surface Extraction from LIDAR|
|Department||Department of Civil and Environmental Engineering||Supervisors||Dr. Ofer Zilberstain|
|Professor Emeritus Yerach Doytsher|
The ever increasing demand for a three-dimensional presentation of urban areas in various areas, such as urban planning and updating of GIS (Geographical Information Systems) systems, requires updated, accurate, dense and reliable digital data. LIDAR has been more and more the selected technology because of the economy in time, money, and manpower. The data provided by LIDAR systems are presented by means of dense and accurate thee-dimensional points with no classification into various objects, such as buildings, trees, roads, terrain, etc. Various methods have been developed for terrain feature extracting. The most common methods in used are the Robust method and the Rectangle method. The Robust method has been developed specifically for forest regions; its application to urban areas causes wrong classification of buildings as land due to the large area of buildings. The Rectangles method has been developed in particular for flat terrains. The objective of the present research is improving the existing methods, for better results. For the robust method, orthogonal polynomials are being used which permit the use of interpolation functions of the polynomial kind, with no restriction on the degree of the polynom. For the Rectangle method, the determination test is adapted for usage in non-flat terrains. The principal objective of the present research is automatic terrain extraction, with no prior knowledge of the DTM, by means of extraction a roads net, and its application for reconstructing the DTM. The current implementation is applicable in urban areas. The algorithm for terrain extraction have been applied to four urban areas which include various kind of complex objects such as bridges, and large buildings that involved shape containing opening roofs etc. The results show that in each method one can find errors in classification while using only one method. For improving the extraction process, the Roads method and the Robust method have been combined. The results show complete success, at least in the qualitative aspect.