Research on the Application of Deep Learning YOLOv3 in Aerial Patrol Inspection of Optical Cable Lines

Abstract

Since the construction of engineering vehicles such as excavators in the vicinity of optical cable lines is the most important cause of the fault of optical cable lines, it is proposed to apply the deep learning YOLOv3 (You Only Look Once version 3) target detection algorithm to the engineering vehicle detection and warning of the aerial inspection image. Based on the Darknet deep learning framework, the rapid target detection of engineering vehicles is realized by making engineering vehicle datasets. The detection speed is 60ms per sheet, and the target recognition AP (Average Precision) value is 0.869, which exceeds the algorithms such as deep learning Faster R-CNN (Faster Region-based Convolutional Neural Network) algorithm and machine learning HOG+LBP+SVM (Histogram of Oriented Gradient + Local Binary Pattern + Support Vector Machine) algorithm. The research results can provide some reference for aerial patrol and inspection and early warning for optical cable lines.