Research on Image Segmentation of Ice-Covered Wires for Overhead Lines
This paper deeply studies the image segmentation technology of ice-covered wires for overhead lines,and proposes a new recognition method based on Mask R-CNN.This method efficiently processes wire images under complex background,and uses the original images obtained by online monitoring system as input data.After accurate calculation of the model,the method can accurately judge the ice-covered state of the wire in the image(with or without ice),and output the target frame of the ice-covered area and the mask of the ice-covered wire.After the comparison experiment with U-Net model,it is found that Mask R-CNN not only significantly reduces the model training time by 90%,but also improves the accuracy by about 10%,fully demonstrating its excellent performance in the field of image segmentation of ice-covered wires in overhead lines.
万方数据
维普
