基于YOLOv3的电网作业人员安全帽佩戴检测

Detection on safety helmet wearing of power grid operators based on YOLOv3

为了有效监测电网作业人员不规范佩戴安全帽行为,提出1种基于YOLOv3的电网作业现场安全帽佩戴检测方法。针对安全帽佩戴规范性问题,构建正确佩戴、不正确佩戴和未佩戴安全帽3种情况下的图像样本库;并利用该数据库对YOLOv3模型进行训练与测试,结合模型参数、样本比例及算法对比分析,开展电网作业人员安全帽佩戴检测算例。结果表明:YOLOv3模型检测精度能够达到92.59%,同时模型每秒可检测15张图片,在复杂作业场景下能够实现有效检测,可为避免电网作业人员安全隐患提供技术参考。

In order to effectively monitor the nonstandard wearing safety helmet behavior of power grid operators,a method for detecting the wearing of safety helmet at power grid operation site based on YOLOv3 was put forward.Regarding the standardization of wearing safety helmet,an image sample library in three cases of correct wearing,incorrect wearing and unwearing safety helmet was constructed.The database was used to train and test the YOLOv3 model,and combining with the model parameters,sample ratios and algorithm comparison analysis,the cases of safety helmet wearing detection of power grid operators were carried out.The results showed that the detection accuracy of the YOLOv3 model could reach 92.59%.At the same time,the model could detect 15 images per second,which can achieve the effective detection in complex operation scenarios,and provide technical reference for power grid operators to avoid the potential safety hazards.