基于改进YOLOv5s的保护屏柜设备检测模型

Protective Panel Detection Model Based on Improved YOLOv5s

针对电力保护屏柜及其内部装置相似度高、分布密集的特点,提出一种基于YOLOv5s模型的保护屏柜检测模型YOLOv5s-PBC。该模型选取轻量级网络PP-LCNet作为主干网络,减少网络参数,提高检测速度;采用加权双向特征金字塔(BiFPN)结构作为特征融合的基本单元,提高模型的感知能力;引入CARAFE上采样算子进行上采样,解决上采样信息丢失问题。实验结果表明,该模型有较好的识别效果,均值平均精度(m AP)达到92.6%,fps达到50.25,模型大小仅为17 MB,适合在移动嵌入式设备上部署,具有一定的实用价值。

In view of the characteristics of high similarity and dense distribution of internal devices in power protective panel,this paper proposes a detection model for protective panel,called YOLOv5s-PBC.The model is based on YOLOv5s and uses a lightweight network called PP-LCNet as the backbone feature extraction network to reduce network parameters and improve detection speed.It uses the weighted bi-directional feature pyramid network(BiFPN)structure as the basic unit for feature fusion to enhance the model's perception ability.The CARAFE upsampling operator is introduced to overcome the problem of information loss during upsampling.The experimental results demonstrate that the model has good recognition performance,with a mean average precision(mAP)of 92.6%,a frame rate of 50.25 fps,and a model size of only 17 MB.It is suitable for deployment on mobile embedded devices and has practical value.