ABSTRACT
This chapter shows that building extraction is treated as a machine learning or data mining problem. It presents a brief introduction to the clustering technique, including the related concepts and terms, and explains the difficulties when this technique is applied to Light Detection and Ranging (LiDAR) data. LiDAR point clouds collected over urban areas consist of points reflected from man-made objects like buildings, cars, etc. and from natural surfaces such as bare earth terrain and trees. The chapter describes the breakline detection that is the separation of planar points with nonplanar points, using the principal component analysis (PCA) technique. The PCA is applied to determine the necessary basis vectors. Data clustering is a widely used technique in pattern recognition, remote sensing, and data mining. Through this technique, a heterogeneous dataset with different properties will be partitioned into a number of subsets such that each subset should share similar properties.
