Volumetrie features for object region classification in 3D LiDAR point clouds

Abstract

LiDAR data is a set of geo-spatially located points which contain (X, Y, Z) location and intensity data. This paper presents the extraction of a novel set of volume and texture-based features from segmented point clouds. First, the data is segmented into individual object regions using an automatic seeded region growing technique. Then, these object regions are normalized to a N × N × N voxel space, where each voxel contains information about the location and density of points within that voxel. A set of volumetric features are extracted to represent the object region; these features include: 3D form factor, rotation invariant local binary pattern (RILBP), fill, stretch, corrugation, contour, plainness and relative variance. The form factor, fill, and stretch provide a series of meaningful relationships between the volume, surface area, and shape of the object. RILBP provides a textural description from the height variation of the LiDAR data. The corrugation, contour, and plainness are extracted by 3D Eigen analysis of the object volume to describe the details of the object's surface. Relative variance provides an illustration of the distribution of points throughout the object. The new feature set is robust, and scale and rotation invariant for object region classification. The performance of the proposed feature extraction technique has been evaluated on a set of segmented and voxelized point cloud objects in a subset of the aerial LiDAR data from Surrey, British Columbia, which was available through the Open Data Program. The volumetric features, when used as an input to an SVM classifier, correctly classified the object regions with an accuracy of 97.5 %, with a focus on identifying five classes: ground, vegetation, buildings, vehicles, and barriers.