Performance of histogram descriptors for the classification of 3D laser range data in urban environments

Abstract

The selection of suitable features and their parameters for the classification of three-dimensional laser range data is a crucial issue for high-quality results. In this paper we compare the performance of different histogram descriptors and their parameters on three urban datasets recorded with various sensors-sweeping SICK lasers, tilting SICK lasers and a Velodyne 3D laser range scanner. These descriptors are 1D, 2D, and 3D histograms capturing the distribution of normals or points around a query point. We also propose a novel histogram descriptor, which relies on the spectral values in different scales. We argue that choosing a larger support radius and a z-axis based global reference frame/axis can boost the performance of all kinds of investigated classification models significantly. The 3D histograms relying on the point distribution, normal orientations, or spectral values, turned out to be the best choice for the classification in urban environments.

A Model-Based Approach for Fast Vehicle Detection in Continuously Streamed Urban LIDAR Point Clouds

“…Second, in crowded scenes the vehicles, pedestrians, trees and street furnitures often occlude each other causing missing or broken object parts in the visible measurement streams. Third, typically by terrestrial laser scanning the point cloud density rapidly decreases as a function of the distance from the sensor [3], which fact may cause strongly corrupted geometric properties of the object appearances, misleading the recognition modules. Further requirements arise for navigation or autonomous driving systems, where the data is continuously streamed from a laser sensor mounted onto a moving platform, and we are forced to complete the object detection and recognition tasks within a very limited time frame.…”

An Integrated 4D Vision and Visualisation System

“…High speed Rotating Multi-Beam (RMB) LIDAR systems, such as the Velodyne HDL-64E sensor, are able to provide accurate 3D point cloud sequences with a 15 Hz refreshing frequency, making the configuration highly appropriate for analysing moving objects in large outdoor environments with a diameter up to 100 meters. However, a single RMB LIDAR scan is a notably sparse point cloud, moreover we can also observe a significant drop in the sampling density at larger distances from the sensor and we also can see a ring pattern with points in the same ring much closer to each other than points in different rings [1]. These properties yield poor visual experiences for the observes, when a raw ( Fig.…”

Fast 3-D Urban Object Detection on Streaming Point Clouds

“…First, the raw sensor measurements are noisy. Second, the point density is uneven: [2] typically in terrestrial LIDAR point clouds the point densities dominate from the direction the measurement is taken, causing strongly corrupted geometric properties of the objects such as missing object parts or deformed shapes. The object detection process is further complicated when the data is continuously streamed from a laser sensor on a moving platform or a mobile robot.…”