A method for perceiving position of live-line workers based on image segmentation and depth estimation

Aiming at the problems of the existing safety monitoring technology for live working, such as insufficient accuracy and susceptibility to the influence of the operating environment, this paper proposes an efficient method for obtaining the body position of live-line workers based on image segmentation and depth estimation. The method firstly obtains the live working video based on UAV monitoring platform equipped with positioning module, gets the body contour of the operator through the image segmentation model, then calculates the depth information of the body contour relative to the camera viewpoint through the depth estimation model, and finally extracts the spatial coordinates of the body contour based on the pixel projection method. The above steps can provide high-precision body position sensing technical support for the calculation of the live working safety distance, and then combine with the laser point cloud information to realize the real-time monitoring of the minimum safety distance or the minimum complex gap. The experimental results show that the error of this method is less than 25 cm, and the sensing time of a single image is less than 0.5 s, which realizes the real-time accurate measurement of the operator's body position.