基于改进的YOLOv5s安全帽佩戴检测算法

Improved YOLOv5s helmet wearing detection algorithm

针对个人防护用具安全帽的防护检测识别需求,现有的人工检测方法费时费力,无法做到实时监测.提出了一种基于YOLOv5s深度学习模型的安全帽检测算法,能够有效识别检测安全帽是否正确佩戴.并通过添加CA注意力机制,重新分配每个空间和通道的权重;以BoT3替代原有的C3模型,作为主干网络;并将CIOU损失函数改为SIOU等方法,改进原有的YOLOv5s模型,提高安全帽检测识别的精度,提高检测速度.实验结果表明,安全帽识别检测的平均精度比原始模型提高了2.2%,识别检测速度提升了19 ms,实现了更准确地轻量高效实时的安全帽佩戴检测.

The protection detection and identification requirements for personal protective equipment and helmets,the existing manual detection methods were time-consuming and labor-intensive,they were unable to achieve real-time monitoring.Therefore,this paper proposed a safety helmet detection algorithm based on deep learning,which can effectively identified and detected whether the safety helmet was correctly worn.And by improving the YOLOv5s model of deep learning,adding CA attention mechanism and reallocating the weights of each space and channel;using BOT3 as the backbone network to replace the original C3 model;changing the CIOU loss function to SIOU and other methods to improve the accuracy of helmet detection and recognition,improved detection speed was improved.The experimental results showed that the average accuracy of helmet recognition and detection was improved by 2.2%compared to the original model,and the recognition and detection speed was improved by 19 ms,more accurate,lightweight,efficient,and real-time helmet wearing detection were achieved.