Visual system for oil sampling robot based on YOLO v5 and OpenCV model

The aging problem of transformer oil in Ultra-High Voltage (UHV) substation is crucial in long-time operation. The transformer oil can be removed and tested regularly to avoid the problem of power system breakdown. This paper proposes a visual system for oil sampling robot based on YOLO v5 and OpenCV model under the specific environment of substation. Firstly, YOLO v5 algorithm is used to identify and locate the object in the local view. OpenCV is used to estimate the distance, size and pose of the object by using modules such as morphological operation and minimum enclosing rectangle. Finally, the positioning information is transmitted to control the movement of the manipulator. Experiments show that the proposed method can recognize, classify and locate objects. The UR5E manipulator and end-effector can be used to grasp and control the oil valve accurately. The average time of one full grasp can reach 2.47s, and the success rate of grasp can reach 98%.