Detection for Pin Defects of Overhead Lines by UAV Patrol Image Based on Improved Faster-RCNN

被引：0
|
作者：
Gu C. [1 ]
Li Z. [1 ]
Shi J. [1 ]
Zhao H. [2 ]
Jiang Y. [2 ]
Jiang X. [1 ]
机构：
[1] Department of Electrical Engineering, Shanghai Jiao Tong University, Shanghai
[2] Guangzhou Bureau of UHV Transmission Company China Southern Power Grid, Guangzhou
来源：
Gaodianya Jishu/High Voltage Engineering | 2020年 / 46卷 / 09期
关键词：
Convolutional neural network; Multi-scale feature fusion; Overhead line; Pin defect detection; UAV patrol image;
D O I：
10.13336/j.1003-6520.hve.20190748
中图分类号：
学科分类号：
摘要：
In order to improve the efficiency of patrolling overhead lines by unmanned aerial vehicle (UAV) and improve the detection rate of pin defects of overhead lines, we put forward a method for detecting pin defects of UAV patrol overhead lines based on the improved Faster-RCNN algorithm. Aiming at the UAV patrol images with large background and small target size, we improved the original Faster-RCNN by the following methods: taking the deep residual network ResNet101 as the pre-feature extraction network, increasing the training image scale, establishing the multi-scale fusion feature by the feature pyramid, and optimizing the initial anchor frame by the K-means algorithm. Meanwhile, experiments were carried out with the actual UAV patrol images. Results show that this method has good detection effect for pin defects in UAV patrol image. The precision rate reaches 93.6%, and the recall rate reaches 89.8% in the test data set. Compared with other existing common objection detection methods, this method has better detection effect and strong generalization ability. © 2020, High Voltage Engineering Editorial Department of CEPRI. All right reserved.
引用
收藏
页码：3089 / 3096
页数：7
相关论文
共 23 条
  • [1] CHEN Qinyin, FAN Jing, Exploration application of UAV in transmission line inspection, Electric Engineering, 40, 3, pp. 80-81, (2019)
  • [2] CHEN Chi, PENG Xiangyang, SONG Shuang, Et al., Safety distance diagnosis of largescale transmission line corridor inspection based on LiDAR point cloud collected with UAV, Power System Technology, 41, 8, pp. 2723-2730, (2017)
  • [3] JIANG Xiuchen, SHENG Gehao, Research and application of big data analysis of power equipment condition, High Voltage Engineering, 44, 4, pp. 1041-1050, (2018)
  • [4] PENG Xiangyang, Qian Jinju, WU Gongping, Et al., Full automatic inspection system and Its demonstration application based on robot for overhead transmission lines, High Voltage Engineering, 43, 8, pp. 2582-2591, (2017)
  • [5] LI Junfeng, WANG Qinruo, LI Min, Electric equipment image recongnition based on deep learning and random forest, High Voltage Engineering, 43, 11, pp. 3705-3711, (2017)
  • [6] HUANG Dongfang, Research of power lines recogni-tion in transmission line image based on an improved hough transform, (2016)
  • [7] WANG Wanguo, ZHANG Jingjing, HAN Jun, Et al., Broken stand and foreign body fault detection method for power transmission line based on unmanned aerial vehicle image, Journal of Computer Applications, 35, 8, pp. 2404-2408, (2015)
  • [8] HUANG Xinbo, LIU Xinhui, ZHANG Ye, Et al., Classification recognition method of insulator in aerial image based on the red-blue difference and developed K-means al-gorithm, High Voltage Engineering, 44, 5, pp. 1528-1534, (2018)
  • [9] HAO Yanpeng, JIANG Xiaolan, YANG Lin, Et al., Evaluation of natural icing condition for in-service insulators based on image segmentation, High Voltage Engineering, 43, 1, pp. 285-292, (2017)
  • [10] GONG Gangjun, ZHANG Shuai, WU Qiuxin, Et al., Foreign body identification of high voltage transmission line based on Tensor Flow, Electric Power Automation Equipment, 39, 4, pp. 204-209, (2019)
123