Abstract
Building clustering is an important task that should be performed prior to building generalization operations. One of the most common approaches for building clustering is the use of density-based algorithms. Current density-based algorithms encounter problems in detecting accurate clusters in a region with varying density. To overcome this problem, a new density-based spatial clustering algorithm, local-adaptive DBSCAN (LA-DBSCAN), which can cluster polygonal buildings in urban blocks with noise and non-uniform density, is developed. The advantage of LA-DBSCAN is that it can select parameters that are adaptive to different local situations. To evaluate the performance of the proposed model, the complete building generalization process is implemented using four datasets at 1:25k scale. An evaluation of the results allowed us to conclude that the LA-DBSCAN algorithm yields more homogeneous and accurate results than the DBSCAN algorithm. Thus, the presented approach is beneficial for the detection of building patterns and the generalization.
Disclosure statement
No potential conflict of interest was reported by the authors.