Design Of Visual System Handed In Tokamak Flexible In-vessel Inspection Robot

First Wall, which is the inner wall of the nuclear fusion reactor Tokamak in-vesselchamber, is the core component of Tokamak device. It is easily damaged, or produce harmfulsubstances in the tough environments. Therefore we require visual inspection system handedin an endoscopic arm to provide accurate site information, so as the bricks can be replaced,repaired, or maintained. In this paper, we design a visual inspection system. It is illustratedas follows:On the one hand, high precision and large range of the inspection are both crucial to thetask. On the other hand, the system requires adaption to a tough environment (hightemperature, the darkness, etc) also. At the same time, a quick and authentic inspection ofthe obtained images of brick surface is required. We completed the binocular visiondetection structure: high-definition industrial camera (detection camera) can be in finescanning mode for inspection, in order to detect damage of brick surface, while IP cameraas a surveillance camera, can be in coarse scanning mode for a quick and wide field-of-viewvideo recording.In order to meet the real-time requirements of the video/image transmission, and reducethe packet loss rate, the special image/video transmission network is developed. Thanks tothe network, multiple hosts receive the image at the same time, with different detection taskdistributed to one another.In addition, positioning of the smallest unit of the first wall is one of the key tasks ofvisual inspection system. One is to develop robust image processing algorithms to detectbrick unit. y. The another is to take advantage of the characteristics of the first wall, andacquire arm coordinates and poses, to match the brick schema quickly.Finally, due to the arm structure characteristics of “firm-soft combination”, the vibration estimate of the end effector is the prior step of vibration suppression. The camerahanded in the end of the arm is used to acquire the image to estimate vibration, which is akind of high efficiency and low cost, especially in the environment with electromagneticinterference. The use of short-time Fourier transform, adding Hanning Window and selectthe adaptive length of window to estimate the vibration parameters, can predict vibrationinformation accurately and filter out the noise.We conduct a series of experiments, including visual communications test, GUI functionaltest and tile positioning algorithm test. What’s more, demonstrate the effectiveness ofvibration estimation methods with the camera mounted on the end of the manipulator, whichis a single-link manipulator. Simulate it several times. Experiment results verify theeffectiveness of the proposed methods.