Abstract
Abstract. The automatic analysis of large 3D point clouds represents a crucial task in photogrammetry, remote sensing and computer vision. In this paper, we propose a new methodology for the semantic interpretation of such point clouds which involves feature relevance assessment in order to reduce both processing time and memory consumption. Given a standard benchmark dataset with 1.3 million 3D points, we first extract a set of 21 geometric 3D and 2D features. Subsequently, we apply a classifier-independent ranking procedure which involves a general relevance metric in order to derive compact and robust subsets of versatile features which are generally applicable for a large variety of subsequent tasks. This metric is based on 7 different feature selection strategies and thus addresses different intrinsic properties of the given data. For the example of semantically interpreting 3D point cloud data, we demonstrate the great potential of smaller subsets consisting of only the most relevant features with 4 different state-of-the-art classifiers. The results reveal that, instead of including as many features as possible in order to compensate for lack of knowledge, a crucial task such as scene interpretation can be carried out with only few versatile features and even improved accuracy.
Highlights
The automatic interpretation of large point clouds with irregularly distributed 3D points is a task of major interest in photogrammetry, remote sensing and computer vision
Instead of using complex models and finding the optimal training technique for learning contextual relationships, we focus on the use of good and versatile features which provide a higher interpretability compared to numerous interactions between a large number of elements
In this paper, we address two important issues: 1) which features can be calculated for each 3D point of a point cloud and 2) how meaningful are these features for further tasks such as 3D scene analysis? Aiming to reach general applicability, we only exploit geometric features for 3D scene analysis and neglect echo-based features as well as full-waveform features which are included for similar approaches (Chehata et al, 2009; Mallet et al, 2011), but not always available
Summary
The automatic interpretation of large point clouds with irregularly distributed 3D points is a task of major interest in photogrammetry, remote sensing and computer vision. Airborne Laser Scanning (ALS) for instance is used for capturing large-scale 3D environments with almost homogeneous point density. The local point density still remains relatively low and reaches up to only about 50 pts/m2. Terrestrial Laser Scanning (TLS) and Mobile Laser Scanning (MLS) are applied for capturing dense and accurate 3D information representing local object surfaces, but the density of the measured 3D points depends on their distance to the scanning unit. An appropriate interpretation of the captured data has to face certain challenges arising from either low or varying point density
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
