Research on Detection Algorithm of Catenary Insulator Based on Improved Faster R-CNN

摘要

The traditional method of regular inspection of the catenary is mainly manual inspection, in order to improve the automation level of catenary detection and solve the problem of intelligent detection of catenary insulator objects in complex environments, this paper figures out an improved Faster R-CNN catenary insulator object detection method. This improved method improves the accuracy of insulator detection by adding a deconvolution layer after the last layer of CONV5 in the VGG16 network to improve the resolution of the deep insulator feature map. The consequence of this research indicates that the improvement method can accurately locate and identify the insulator in the complex catenary image, and has good universality and high detection efficiency for the shooting angle and shooting distance.