Assisting Wind Turbine Hoisting with Yolov7 and Object Tracking Technology
Chenzhe Ma, Bo Fang, Hao Liu, Shun Li, Minmin Ma · 2023
In the process of hoisting wind turbines, the driver of the crane cannot observe the panoramic view of the components being hoisted and is completely dependent on the scheduling of the commander, which restricts the efficiency and safety of hoisting. We propose an assisted hoisting algorithm for wind turbines. A high-definition camera is installed at the hoisting site to capture the hoisting scene in real time, and then Yolov7/ Yolov7-tiny is used to detect the object being hoisted in the frame. The box of the object is used as the input of the Deepsort algorithm to track the hoisting object in real time. Finally, the tracking information is used to adjust the angles of the pan and tilt and the focal length of the camera to provide the crane driver with a clear hoisting panorama. Experiments and engineering applications show that the mAP of Yolov7 reaches 0.88(Yolov7-tiny reaches 0.81), and the real-time tracking speed exceeds 12 FPS, which meets the requirements of use.