3D vision and object recognition for off-shore robot guidance
The Norwegian oil company StatoilHydro has developed a new concept for an unmanned oil & gas platform. The platform should be equipped with intelligent and reliable robotic and instrumentation systems that enable a series of inspection and maintenance tasks to be carried out either automatically or remotely controlled from operators located onshore.
In order to support research on robotic and instrumentation systems for this platform concept, StatoilHydro has financed a robotic lab facility in Trondheim. The lab has been developed by the GEMINI Centre for Advanced Robotics, where both SINTEF and NTNU participate.
One example of a maintenance task that has been demonstrated in the lab is to change the battery of a sensor with a robot. Since we cannot assume that the accurate position of the sensor is known in advance, the robot has been equipped with 3D vision based on structured light
Advanced 3D object recognition techniques are used to detect the position and orientation of the battery based on a 3D image of the process equipment and a CAD model of the battery.
The image above to the left shows the process equipment with a magnified image of the sensor battery that we want to change. The image to the right displays a 3D scan of this area captured by a structured light instrument on the robot. 3D points that are found to reside on the surface of the battery are indicated in green, while the detected battery has been annontated in red on the 3D image.
Left: A magnified region of the 3D image with measured points on the surface of the battery indicated in green. The CAD model of the battery is annotated in red in the detected position and orientation. Right: The position and orientation of the battery is sent to a robot with a suitable gripper tool, which is instructed to replace the battery with a new one.
Video

Click the image above to view a video that demonstrates the entire operation of detecting and replacing a sensor battery by equipping the robot with 3D vision and sophisticated 3D object recognition algorithms.
Related projects
For more information, please contact Øystein Skotheim.