Efficient camera calibration and edge detection of robot vision system for ITER blanket remote handling system

Maruyama, Takahito; Noguchi, Yuto; Ueno, Kenichi; Takeda, Nobukazu; Kakudate, Satoshi

In the vacuum vessel of ITER, the blanket modules will be handled by a huge robotic manipulator that travels on a rail which has been deployed into the vessel. This system is called the ITER blanket remote handling system. A robot vision system using two cameras was developed for rough positioning of that system. Previous testing confirmed that the manipulator satisfies the required positioning accuracy with that robot vision system. However, these tests also showed that camera calibration and edge detection take quite some time since precise camera positioning is needed to calibrate the cameras, and parameters need to be adjusted during operations for edge detection. To improve camera calibration, software image transformation to correct the misalignment of cameras was adopted. Using this method, positioning of cameras does not affect accuracy of the robot vision system, and thus precise positioning of cameras is not necessary. To improve edge detection, we adopted smaller apertures for a wider depth of field, edge dilation for connecting edges, and the Sobel operator for detecting the edges of the modules. These methods make edge detection robust and the adjustment of parameters unnecessary. Through testing we confirmed that these methods make the robot vision system more efficient.