Obstacle Detection using Structured Light
This research has been folded into the EXPLORING THE WORLD WITH A RAY OF LIGHT, as a dual use problem for identifying obstacles for the visually impaired as well as mapping the environment for a ground mobile robot.
A variety of laser based scanners and distance sensors have been used in computer vision for object detection in robotics applications. Most of these systems are expensive and require complex calibration techniques and regular maintenance. We propose to construct a low-cost distance and object mapping system constructed from common parts and interfacing to a regular PC.
Multiple laser based triangulation devices have been developed to determine object distance information in a camera coordinate system. Most require complex calibration techniques and depend on sensitive video cameras and video capture electronics working in conjunction with a high performance laser. Others use a time-of-flight (TOF) based system that includes moving parts and are also expensive and require regular maintenance.
The goal of our project is to expand on these existing technologies by adapting them so they can be implemented with minimal monetary cost and a computational complexity that is low enough that they can be implemented in consumer robotics applications. The biggest challenge in the development of consumer robotics currently is to make sensors available that are cheap enough for mass production and yet are still reliable for autonomous use. The goal of this project is to promote the development of a distance sensor that is robust yet is very inexpensive and easy to calibrate.
This laser striping triangulation device was designed to fulfill several important characteristics. First, the device should be eye safe, such that it does not need to be isolated from people while it is operating. Second, to reduce overall system price it should be constructible using commercially available parts. Third, it should use the camera’s resolution as effectively as possible to provide a wide range of angles. The device should be able to detect distances within two meters in normal interior lighting conditions with accurate resolution. Finally, it should be capable of operating at a useful acquisition rate (around 10Hz).
multimedia
Publications
-
(1)Ilstrup, D., Elkaim, G., “Low Cost, Low Power Structured Light Based Obstacle Detection,” ION/IEEE Position, Location, and Navigation Symposium, ION/IEEE PLANS 2008, Monterey, CA, May 5-8, 2008, pp. 865-870 (pdf)
-
(2)Ilstrup, D., Elkaim, G., “Single Frame Processing for Structured Light Based Obstacle Detection,” ION National Technical Meeting, ION NTM 2008, San Diego, CA, Jan. 28-30, 2008, pp. 514-520 (pdf)
People
-
• Roberto Manduchi, Professor, Computer Engineering, UCSC, 831.459.1479
-
• Gabriel Elkaim, Associate Professor, Computer Engineering, UCSC, 831.459.3054
-
• David Ilstrup, PhD student (graduated), CE UCSC, now at Honda Research
