基于改进YOLOv5的安全帽检测算法

Helmet Detection Algorithm Based on Improved YOLOv5

安全帽作为劳动者最基本的保护,对劳动者的生命安全具有重要意义。但是,由于缺乏安全意识,在建筑工地中存在安全帽佩戴不规范的情况。随着目标检测技术的不断发展,高精度、高效率的YOLO系列算法已经被应用于各种场景检测任务中。为建立起数字化的安全帽监测系统,该文首先对建筑工地中安全帽佩戴情况所采集的7 581张图片进行标注。然后,提出一种基于改进的YOLOv5的安全帽检测方法,并使用不同参数的YOLOv5(s,m,l)模型进行训练和测试,对这3种模型进行比较和分析。使用可训练目标探测器YOLOv5s的m AP达到91.7%,证明基于改进的YOLOv5的头盔探测的有效性。

As the most basic protection of workers, safety helmets are of great significance to the life and safety of workers.However, due to the lack of safety awareness, there is a non-standard wearing of safety helmets at construction sites. With the continuous development of target detection technology, high-precision and efficient YOLO algorithms have been applied to a variety of scene detection tasks. In order to establish a digital safety helmet monitoring system, this paper first marks 7 581 pictures collected from the wearing of safety helmets in the construction site. Then, a helmet detection method based on improved YOLOv5is proposed, and YOLOv5(s,m,l) models with different parameters are used for training and testing, and the three models are compared and analyzed. The mAP using trainable target detector YOLOv5s reaches 91.7%, which proves the effectiveness of helmet detection based on improved YOLOv5.