Robot navigation in pipelines using 3D cameras
Robots used for inspection of pipeline systems, such as sewer systems and gas pipelines, are in need of a navigation system. Current pipe inspection robots use stereo cameras or ultrasound for navigation, giving sparse datasets with a high degree of uncertainty. SINTEF is developing a robot navigation system that uses a 3D camera for junction-, bend- and obstacle detection, providing continuous and accurate position estimates.
Time-of-flight (TOF) cameras provide 3D information at video frame rates by determining both range and intensity at each pixel. The cameras are active sensors that measure the time taken by infrared light to travel to the object and back to the camera, and they have a big potential in a wide range of application areas.
SINTEF uses a TOF camera mounted in front of the pipe robot for navigation. The camera provides dense 3D information from the surroundings in front of the robot, which enables a continuous and accurate estimation of junctions and bends in the pipeline. Detection of such landmarks in 3D images is done using algorithms developed by SINTEF in this and similar projects.
Figure 1: Time-of-flight camera image from inside the pipeline. Pixel color indicates range. A junction is discernible to the right in the image, at one meter distance.
Figure 1 shows an example of a range image taken with a TOF camera inside the pipeline. A junction is discernable to the right in the image. Combined with a coarse map of the pipeline system, the detected landmarks can be used for estimation of current robot position and further navigation in the pipe. Additional sensor information, such as odometry and inertial measurements, will be used for validation of the position estimates.
Figure 2: The pipe inspection robot PiKo implements the navigation system based on a 3D camera.
Figure 2 shows the pipe inspection robot PiKo, developed in cooperation with the Department of Applied Cybernetics at SINTEF. This is a new type of robot for inspection of complex pipe systems, including vertical pipelines and pipes of varying diameters. The navigation system based on a TOF 3D camera and additional sensors will be implemented on PiKo.
Related article:
Pipeline landmark detection for autonomous robot navigation using time-of-flight imagery
Thielemann, J.T.; Breivik, G.M.; Berge, A.
Computer Vision and Pattern Recognition Workshops, 2008. CVPR Workshops 2008. IEEE Computer Society Conference on
Volume, Issue, 23-28 June 2008 Page(s): 1 - 7
For further information please contact Jens T. Thielemann.
