The Use of Random Forest for the Classification of Point Cloud in Urban Scene

The aim of the paper concerns the classification of the point cloud using a suitable method based on the Random Forest algorithm. In addition, thanks to the use of specific sensors, such as Airborne Laser Scanning (ALS), it is possible to obtain a georeferenced point cloud in a short time, which makes it possible to represent and model urban areas not only through a graphic representation but also through a semantic one. The development of a suitable methodology made it possible to automatically classify a point cloud acquired on an urban scene acquired with a "Leica City Mapper"sensor over the city of Bordeaux (France). The quality of the point cloud classification was evaluated using appropriate performance indices (Overall Accuracy and F1 measure), which showed encouraging results on the quality of the developed method.