Terrain Classification in Complex Three‐dimensional Outdoor Environments

Abstract

This paper presents two techniques to detect and classify navigable terrain in complex 3D environments. The first method is a low level on-line mechanism aimed at detecting obstacles and holes at a fast frame rate using a time-of-flight camera as the main sensor. The second technique is a high-level off-line classification mechanism that learns traversable regions from larger 3D point clouds acquired with a laser range scanner. We approach the problem using Gaussian Processes as a regression tool, in which the terrain parameters are learned, and also for classification, using samples from traversed areas to build the traversable terrain class. The two methods are compared against unsupervised classification, and sample trajectories are generated in the classified areas using a non-holonomic path planner. We show results of both the low-level and the high-level terrain classification approaches in simulations and in real-time navigation experiments using a Segway RMP400 robot.

ToF cameras for active vision in robotics

“…If we set the focus to obtain sharp images of closer objects then the depth of field is small. ToF cameras do not have auto-focus 1 It is commonly accepted that 0.7m is the closest distance, but in our tests we have been able to obtain depth images at 0.5m. New Kinect camera, to appear in the beginning of 2014 is supposed to work at 0.3m.…”

“…A radio frequency modulated light field is emitted and then reflected back to the sensor, (a) Scene-related task setup: ToF camera attached at the front of a mobile robot used for navigation and terrain classification. Reproduced from [1] (b) Object-related task setup: ToF camera attached to a manipulator end-effector, in this case a Chlorophyll meter, to get measures by touching the leaf surface. which allows for the parallel measurement of its phase (crosscorrelation), offset, and amplitude [3].…”

“…In robotics it has been used for a wide range of applications that can be classified in two groups: scene-related and object-related. In scene-related tasks usually the camera is mounted on a mobile robot (Fig 1a) and it is used for mapping and localization, although it has been also applied to obstacle detection and terrain classification [1]. In object-related tasks, the camera is usually attached to the end-effector of a robot manipulator (eye-in-hand configuration), so that new images can be obtained by actively changing the point of view of the camera (Fig.…”

“…In each figure can be observed: (1)robot scene, (2)ToF intensity image, (3)ToF pointcloud segmented (green points represent horizontal planes, red points vertical planes), (4)ToF pointcloud rotated. Reproduced from [1]. 10 7 Details of two different tools in the end-effector of (a) a WAM robot and (b) a Kuka Lightweight robot.…”

Building Fuzzy Elevation Maps from a Ground-Based 3D Laser Scan for Outdoor Mobile Robots

“…Environment mapping is a key issue in mobile robots targeted to unstructured terrains [1,2]. In this sense, three-dimensional (3D) laser scans provide valuable information for applications such as planetary exploration [3][4][5] or urban search and rescue [6,7].…”