为解决施工过程中操作不规范导致安全事故频繁发生的问题,研究了一种基于 Yolov5 的施工现场异常行为智能检测方法.该方法能够检测施工工人是否佩戴安全帽,并监测大型设备驾驶人员是否有抽烟、打电话等不当行为.针对原模型对小目标检测能力不佳问题,通过改进 YOLOv5 s 算法,使用 Focal-EIoU 损失函数来解决 YOLOv5s中CIoU Loss计算回归结果不准确的问题,从而提高模型对小型目标的检测精度,实现对检测方法的优化.实验结果验证了该方法在不同场景下具有良好的性能,有助于有效提升施工现场的安全管理水平.
YOLOv5-based Abnormal Behavior Detection at Construction Sites
In view of frequent safety accidents caused by improper operations during construction,the present work stud-ied a Yolov5-based intelligent detection method for abnormal behaviors at construction sites.This method can detect whether construction workers are wearing safety helmets and detect improper behaviors of large vehicle operators such as smoking and making phone calls.To address the problem of the original model's poor ability to detect small targets,the YOLOv5s algorithm was modified by using the Focal-EIoU loss function to solve the problem of inaccurate regression re-sults of the CIoU loss calculation in YOLOv5s,thereby improving the model's detection accuracy of small targets and op-timizing the detection method.Experimental results verified that this method has good performance in different scenarios and facilitates effective improving safety management level of construction sites.
万方数据
维普
