Detection and Classification of Pole-like Objects from Mobile Laser Scanning Data of Urban Environments

The Mobile Laser Scanning (MLS) system can acquire point clouds of urban environments including roads, buildings, trees, lamp posts etc. and enables effective mapping of them. With the spread of the MLS system, the demands for the management of roads and facilities using MLS point clouds have increased. Especially, pole-like objects (PLOs) such as lamp posts, utility poles, street signs etc. are strongly expected to be managed efficiently. We propose a method for detecting PLOs from MLS point clouds and classifying them into three classes: utility poles, lamp posts, and street signs. Our detection method is based on the feature extraction using point classification by Principal Component Analysis (PCA). On the other hand, our classification method is based on not only shape features of the PLOs, but also context features which are derived from the surrounding PLOs distributions. In order to evaluate the accuracy of PLOs detection and classification through our method, we applied our method to MLS point clouds of urban environments.