FEATURE ANALYSES FOR GEODATABASE MAINTENANCE USING AERIAL IMAGERY AND LIDAR DATA

摘要

The maintenance of city geodatabase is an essential work. Parts of buildings may be removed or rebuilt in different periods so that the geodatabase should be timely updated. However, to renew all building models in a city scale is costly and time consuming. Change detection of building models for maintenance, thus, becomes a valuable alternative. In the geospatial area, airborne LIDAR data provides accurate elevation information, and aerial imagery shows rich 2D features. This investigation integrates the latest LIDAR data and aerial imagery to detect changed buildings for further reconstruction processes. Considering the geometric displacements between these two heterogeneous datasets, data registration should be done in advance. Traditional methods modified these local displacements through selected corresponding features. For the improvement of automation, this investigation implements a feature-based method to enhance data registration and change detection at the same time. The proposed scheme contains three steps: (1) feature detection, (2) data registration, and (3) change detection of existing building models. The first step detects linear features from aerial imagery and LIDAR data. These detected features provide the geometrical clues for the detection of building changes. The second step selects unchanged buildings for data registration. We also employ these features to compare with building models and fine-tune the registration results. The last step estimates changed buildings, which have different structure lines between models and used datasets. The preliminary results indicated that the feature-based concept could enhance the procedures of data registration and change detection with better performance.