Building Extraction from LiDAR Using Edge Detection




Miller, Justin

Journal Title

Journal ISSN

Volume Title



Light Detection and Ranging (LiDAR) has become a versatile data source for many applications including building detection. Previously, manual photogrammetric methods were needed to accurately digitize building footprints, often resulting in ineffective and costly data collection process. Automated building extraction from imagery has been studied extensively, in particular using edge detection and image segmentation methods. However, the utilization of such methods, and in particular edge detection, for building extraction from LiDAR has not been fully explored. Consequently, this research explores the use of edge detection-based feature extraction as a possible framework for building detection and delineation in LiDAR data. In particular, building on existing edge detection and image segmentation operators, the proposed framework utilizes a rotating kernel for detecting the edges of buildings as well as the watershed segmentation operator for segmenting and identifying each building. Once identified, each building is then delineated using a combination of Hough transform and topological polygon construction, resulting in the building’s footprint. The building extraction process was tested on three different datasets containing buildings of various shapes, and the extraction results were compared to manually extracted footprints in order to evaluate the accuracy and precision of the proposed framework. The analysis results show that the proposed framework was 90% accurate with 95% of the extracted results area overlapped the manually extracted footprints.



LiDAR, Building extraction, Hough transform, Edge detection