变电站周边环境的红外目标检测

Infrared target detection in the surrounding environment of Substation

针对当前国内变电站周边环境巡检的局限性,基于无人机和红外成像技术,提出了一种改进的YOLOv5红外图像检测方法.通过将主干网络替换成MobileNetV3轻量化网络,来减少参数量;通过修改特征融合层、调整边界框损失函数来增强网络对红外小目标的检测能力.最后通过对自建红外目标检测数据集进行目标检测,并进行对照实验,来验证本文改进算法的可行性.实验结果表明改进网络降低了网络参数,简化了网络结构,对于提高变电站周边环境的监控效率具有参考价值.

In view of the limitations of current domestic substation surrounding environment inspection,an improved YOLOv5 infrared image detection method is proposed based on UAV and infrared imaging technology Reduce the a⁃mount of parameters by replacing the backbone network with MobileNetV3 lightweight network;The detection ability of the target is enhanced by modifying the boundary function of the infrared fusion layer Finally,the feasibility of the improved algorithm is verified by target detection on the self-built infrared target detection data set and compara⁃tive experiments The experimental results show that the improved network reduces the network parameters and sim⁃plifies the network structure,which has reference value for improving the monitoring efficiency of the surrounding environment of the substation.