基于轻量化网络的变电站缺陷图片检测算法
    点此下载全文
引用本文:琚泽立1,邢伟2,金鸿鹏3,徐方植3,蒲路1,侯喆1.基于轻量化网络的变电站缺陷图片检测算法[J].电网与清洁能源,2020,36(8):43~49
摘要点击次数: 178
全文下载次数: 75
作者单位
琚泽立1 1. 国网陕西省电力公司电力科学研究院 
邢伟2 2. 国网陕西省电力公司 
金鸿鹏3 3.西安交通大学 电气工程学院 
徐方植3 3.西安交通大学 电气工程学院 
蒲路1 1. 国网陕西省电力公司电力科学研究院 
侯喆1 1. 国网陕西省电力公司电力科学研究院 
基金项目:陕西省重点研发计划-国际科技合作计划项目(2019KW-010)
中文摘要:保证变电站运行安全和可靠,自动检测变电站设备缺陷是电网智能化的关键。提出基于轻量化网络的变电站缺陷图片检测算法。针对目前传统的深度网络深度较深,耗时较长,易发生梯度爆炸、消失的情况,以及变电站智能监控实时性要求高的问题,提出了一种基于轻量化网络的变电站缺陷图片检测算法。该算法通过深度网络的轻量化设计,不仅可降低计算时间复杂度和空间复杂度,并且能提升检测准确度。该算法利用轻量化的特征提取网络进行图片的多尺度特征提取,且能根据多尺度特征进行目标种类和位置的检测。以变电站现场故障图片作为实验数据,对比测试了SSD算法和所提算法在面对不同故障的检测准确率。实验结果表明,所提算法在变电站故障检测任务中比SSD算法更加准确,为变电站故障检测提供了有效的手段。
中文关键词:缺陷图片检测  变电站  目标检测  轻量化网络
 
Detection of Substation Faults Image Based on Lightweight Network
Abstract:Automatic detection of substation equipment defects is the key to power grid intelligence to ensure the safety and reliability of substation operation. In this paper, an algorithm based on lightweight network is proposed to detect the defect image of a substation. At present, the traditional deep network is deep, time-consuming and prone to gradient explosion and disappearance, and requires high computing power. In view of the weak computing power and high real-time requirements of the intelligent monitoring equipment in substations, the lightweight algorithm proposed in this paper can reduce the time and space complexity of deep network computing, and improve the detection accuracy. The lightweight feature extraction network is used to extract the multi-scale features of the image, and the detection of the type and position of the target is also carried out according to the multi-scale features. The field fault pictures of the substation are used as experimental data to test the detection accuracy of both the SSD algorithm and the proposed algorithm in the face of different faults. The experimental results show that the algorithm proposed in this paper is more accurate than the SSD algorithm in the substation fault detection, which provides an effective means for substation fault detection.
keywords:defect image detection  substation  object detection  lightweight network
查看全文  查看/发表评论  下载PDF阅读器
    《电网与清洁能源》杂志

期卷浏览

关键词检索

最新公告栏

您是第3521975位访问者    Email: psce_sn@163.com
版权所有:电网与清洁能源网 陕ICP备16002958号-1
本系统由北京勤云科技发展有限公司设计