Automated Parcel-Based Building Change Detection using Multitemporal Airborne LiDAR Data

This paper presents an automated method for parcel-based building change detection using multitemporal light detection and ranging (LiDAR) data. Building footprint polygons were extracted from LiDAR data and simplified using mathematical morphological operations (MMO) to eliminate edge effects. The slope and volume properties of objects in individual parcels were derived from normalized digital surface models (nDSM) and differenced digital surface models (dDSM), and four types of building change were automatically determined. The method was applied to multitemporal LiDAR data in the City of Surrey, British Columbia (Canada) collected in March 2009 and April 2013. Quantitative assessments of the results from 1,112 buildings using measures of completeness, correctness, and quality suggest that the method performed very well in the study area. Causes of errors are analyzed, and possible improvements are discussed. It is concluded that the method for multitemporal LiDAR data can potentially be used for efficient and effective detection of building change associated with land use/land cover change and disaster damage assessment without support from additional remotely sensed data.