针对垃圾数量繁多及一张图片包含多个垃圾物体的情况,文中提出基于改进MobileNet网络的垃圾检测与分类算法。将MobileNet网络融合进YOLOv5(You Only Look Once v5)目标检测算法,同时在主干部分引入卷积注意力模块(Convolutional Block A...针对垃圾数量繁多及一张图片包含多个垃圾物体的情况,文中提出基于改进MobileNet网络的垃圾检测与分类算法。将MobileNet网络融合进YOLOv5(You Only Look Once v5)目标检测算法,同时在主干部分引入卷积注意力模块(Convolutional Block Attention Modul,CBAM)筛选有意义的信息,利用视觉Transformer聚合形成图像特征,并加入使用了加权双向特征金字塔网络区别不同特征的贡献度,引入高效通道注意力(Efficient Channel Attention,ECA)模块对图像特征进行组合并传递给预测层。最后,为了在垃圾目标之间有遮挡的情况下获得更好的性能,使用软性非极大值抑制(soft-Non Maximum Suppression,soft-NMS)方法,并利用Alpha-IoU(Alpha-Intersection over Union)损失函数对提取的特征进行预测。实验结果表明,所提方法能够实现多目标多类别垃圾的定位与识别,mAP(mean Average Percision)值达到了90.31%,相较于YOLOv5网络提升了4.95%,处理速度缩短了约2.4 s。相较于融合ResNet(Residual Network)网络的Faster R-CNN(Region-based Convolutional Neural Network)算法,所提算法在保证准确率的前提下提升了处理效率。展开更多
This paper presents a three-phase integrated power electronic transformer(PET)topology with the capability of DC fault clearance based on the investigation of PET topology.The proposed PET has three ports of high-volt...This paper presents a three-phase integrated power electronic transformer(PET)topology with the capability of DC fault clearance based on the investigation of PET topology.The proposed PET has three ports of high-voltage AC,high-voltage DC and low-voltage DC,which can achieve the flexible access of distributed energy resources・The power unit of the PET adopts a three-phase integrated topology,and the single-phase fluctuating power in the DC-link capacitor is reduced by a 4-port DC/DC converter,thereby reducing the capacitance of the DC-link capacitor;Meanwhile,the power unit uses the clamp double sub-module(CDSM)to allow for self-clearing of the short-circuit fault on the high-voltage DC side.Finally,this paper verifies the proposed PET through an EMT simulation and experimental prototype.展开更多
文摘针对垃圾数量繁多及一张图片包含多个垃圾物体的情况,文中提出基于改进MobileNet网络的垃圾检测与分类算法。将MobileNet网络融合进YOLOv5(You Only Look Once v5)目标检测算法,同时在主干部分引入卷积注意力模块(Convolutional Block Attention Modul,CBAM)筛选有意义的信息,利用视觉Transformer聚合形成图像特征,并加入使用了加权双向特征金字塔网络区别不同特征的贡献度,引入高效通道注意力(Efficient Channel Attention,ECA)模块对图像特征进行组合并传递给预测层。最后,为了在垃圾目标之间有遮挡的情况下获得更好的性能,使用软性非极大值抑制(soft-Non Maximum Suppression,soft-NMS)方法,并利用Alpha-IoU(Alpha-Intersection over Union)损失函数对提取的特征进行预测。实验结果表明,所提方法能够实现多目标多类别垃圾的定位与识别,mAP(mean Average Percision)值达到了90.31%,相较于YOLOv5网络提升了4.95%,处理速度缩短了约2.4 s。相较于融合ResNet(Residual Network)网络的Faster R-CNN(Region-based Convolutional Neural Network)算法,所提算法在保证准确率的前提下提升了处理效率。
基金supported by National Key Research and Development Program of China(2016YFB0400505)Science and Technology Project of State Grid Corporation(SGRIGLBDTKJ[2017]676).
文摘This paper presents a three-phase integrated power electronic transformer(PET)topology with the capability of DC fault clearance based on the investigation of PET topology.The proposed PET has three ports of high-voltage AC,high-voltage DC and low-voltage DC,which can achieve the flexible access of distributed energy resources・The power unit of the PET adopts a three-phase integrated topology,and the single-phase fluctuating power in the DC-link capacitor is reduced by a 4-port DC/DC converter,thereby reducing the capacitance of the DC-link capacitor;Meanwhile,the power unit uses the clamp double sub-module(CDSM)to allow for self-clearing of the short-circuit fault on the high-voltage DC side.Finally,this paper verifies the proposed PET through an EMT simulation and experimental prototype.
