Building Corner Feature Extraction Based on Fusion Technique with Airborne LiDAR Data and Aerial Imagery

Abstract

Generally, automatic building corner or linear feature extraction from urban area aerial imagery is based on traditional computer vision corner or edge detection techniques. However, challenges and difficulties remained due to the complex characteristic of objects in urban images. Visually, the linear features in airborne LiDAR are much more distinct than those in aerial imagery, however, common criticisms arising from the low horizontal accuracy of LiDAR data. To overcome these difficulties, this study proposes a building corner extraction algorithm based on information fusion technology by integrating aerial imagery and airborne LiDAR data. According to experiment results, the proposed method can obtain the distinct building corners not only with the characteristics of uniform spatial distributed pattern based on Voronoi graph theory, but also with the shape, length, and height constrained conditions derived from LiDAR linear features. The proposed algorithm resolves the heterogeneous remote sensing data registration difficulties between LiDAR data and raw aerial imagery.