| 引用本文: | 何宁辉,王世杰,刘军福,等.基于深度学习的航拍图像绝缘子缺失检测方法研究[J].电力系统保护与控制,2021,49(12):132-140. |
| HE Ninghui,WANG Shijie,LIU Junfu,et al.Research on infrared image missing insulator detection method based on deep learning[J].Power System Protection and Control,2021,49(12):132-140 |
|
| 摘要: |
| 为解决目前人工处理分析无人机巡检图像效率低、检测结果受人为因素影响较大的问题，提出了一种图像识别的绝缘子缺失识别方法。首先，对无人机拍摄的图像样本进行了处理，扩充样本集。其次，搭建了绝缘子的检测模型，完成各层网络结构的选择和设计，使用CNN算法实现对绝缘子缺失的检测。随后，构建了绝缘子检测网络，并对各层检测网络参数进行配置。选择实际拍摄的图像作为训练样本进行网络训练。检测结果证实几个指标均在0.95以上，说明算法可准确识别出绝缘子。最后，利用CNN算法对航拍绝缘子进行缺陷检测。绝缘片缺失缺陷的正确识别率为86%。算法可根据检测结果自动显示绝缘子有无缺失缺陷。 |
| 关键词: 绝缘子缺陷 图像处理 Faster R-CNN算法 模型训练 |
| DOI：DOI: 10.19783/j.cnki.pspc.200950 |
| 分类号: |
| 基金项目:宁夏自然科学基金项目资助(2018AAC03222) |
|
| Research on infrared image missing insulator detection method based on deep learning |
|
HE Ninghui, WANG Shijie, LIU Junfu, ZHANG Hao, WU Liangfang, ZHOU Xiu
|
|
(1. State Grid Ningxia Electric Power Research Institute, Yinchuan 750011, China; 2. State Grid Ningxia Electric Power Co., Ltd.,
Yinchuan 750011, China; 3. Zhongwei Power Supply Company, State Grid Ningxia Electric Power Co., Ltd., Zhongwei
751700, China; 4. Yinchuan Power Supply Company, State Grid Ningxia Electric Power Co., Ltd., Yinchuan 750011, China)
|
| Abstract: |
| There is a problem of low efficiency of manual processing and analysis of UAV inspection images, coupled with the large influence of human factors on detection results. Thus a missing insulator recognition method based on image recognition is proposed. First, the image samples taken by the UAV are processed to expand the sample set. Secondly, an insulator detection model is built to complete the selection and design of the network structure at each layer, and a CNN algorithm is used to detect the missing insulators. Subsequently, an insulator detection network is constructed, and parameters of each layer of the detection network are configured. Images actually taken are selected as training samples for network training. The test results have confirmed that several indices are above 0.95, indicating that the algorithm could accurately identify insulators. Finally, the CNN algorithm is used to perform missing insulator detection on aerial insulators. The correct identification rate of missing defects in the insulation sheet is 86%. The algorithm could automatically display whether the insulators have missing defects according to the detection results.This work is supported by the Ningxia Natural Science Foundation (No. 2018AAC03222). |
| Key words: insulator defect image processing Fast R-CNN algorithm model training |
