电子测量与仪器学报

李鑫,刘帅男,杨桢,王珂珂.基于改进 Cascade R-CNN 的输电线路多目标检测[J].电子测量与仪器学报,2021,35(10):24-32

基于改进 Cascade R-CNN 的输电线路多目标检测

Multi-target detection of transmission lines based on improved cascade R-CNN

DOI：

中文关键词: 输电线路多目标检测 Cascade R-CNN 深度学习特征融合

英文关键词:multi-objective identification of transmission lines Cascade R-CNN deep learning feature fusion

基金项目:辽宁省教育厅科学研究经费项目（LJ2019JL013，LJ2020JCL020，LJ2019QL011）、辽宁工程技术大学学科创新团队（LNTU20TD 29）项目资助

作者	单位
李鑫	1. 辽宁工程技术大学电气与控制工程学院
刘帅男	1. 辽宁工程技术大学电气与控制工程学院
杨桢	1. 辽宁工程技术大学电气与控制工程学院
王珂珂	2. 南京电子技术研究所

Author	Institution
Li Xin	1. Faculty of Electrical and Control Engineering, Liaoning Technical University
Liu Shuainan	1. Faculty of Electrical and Control Engineering, Liaoning Technical University
Yang Zhen	1. Faculty of Electrical and Control Engineering, Liaoning Technical University
Wang Keke	2. Nanjing Institute of Electronic Technology

摘要点击次数: 672

全文下载次数: 12643

中文摘要:

针对无人机巡检图像中小目标难以检测、障碍物遮挡目标、正负样本不平衡等问题,提出基于改进 Cascade R-CNN 的输电线路多目标检测方法。改进了 Cascade R-CNN 的特征提取网络,基于 ResNet101 基础网络结构,设计 6 层新型特征金字塔网络(FPN)与之融合,提高了对小目标、重叠目标的检测能力;引入了高斯形式的软非极大值抑制(Soft-NMS)方法,降低了受遮挡影响的目标的漏检率;利用 Focal 损失改进损失函数,缓解了正负样本不平衡对检测精度的影响。训练过程中,基于添加噪声、亮度变换、尺度放缩等数据增强方法扩充数据集,提升了训练模型的泛化性能。实验结果表明,改进的模型在复杂背景下能够对 3 种瓷质绝缘子、瓷质绝缘子缺陷、相间棒、防震锤以及鸟窝同时检测,平均精度均值(mAP)达到 94. 1%,为输电线路的智能巡检提供了一种新思路。

英文摘要:

Aiming at the problems of difficulty in detecting small targets in UAV inspection images, obstacles blocking targets and imbalance between positive and negative samples, a multi-target detection method based on improved Cascade R-CNN was proposed for transmission lines. The feature extraction network of the Cascade R-CNN was improved. Based on the basic network structure of ResNet101, a new 6-layer feature pyramid network structure was designed to achieve feature fusion, improving the detection ability of small targets and overlapping targets. The Gaussian Soft-NMS method was introduced to reduce the missed detection rate of the target with occlusion. The Focal loss was used to improve the loss function, alleviating the impact of imbalance between positive and negative samples on detection accuracy. During the training process, the data set was expanded based on data enhancement methods such as adding noise, brightness transformation and scaling, which improved the generalization performance of the training model. Experimental results show that the improved model can simultaneously detect three types of porcelain insulators, porcelain insulator defects, interphase rods, anti-vibration hammers and bird’ s nests under complex backgrounds. The average accuracy ( mAP) reaches 94. 1%, which provides a new idea for the intelligent inspection of transmission lines.

查看全文查看/发表评论下载PDF阅读器