基于改进YOLOv3的安全帽佩戴检测算法

Research on Helmet Wear Identification Based on Improved YOLOv3

针对复杂工业场景下安全帽佩戴检测存在检测精度低、误检率和漏检率高以及检测速度慢等问题,提出一种改进YOLOv3的识别精度高、检测速度快的安全帽佩戴检测算法。对传统YOLOv3主干网络进行裁剪改进,使检测速度得到明显提升;引入空间金字塔池化模块使局部特征和全局特征更有效地融合;将损失函数改进为CIoU以提升目标预测框与真实目标框的拟合效果;扩充第四特征融合尺度用于小目标检测以提高小目标的识别精度。结果表明:在复杂工业环境下,改进后的YOLOv3安全帽佩戴检测的平均检测精度提高了2.37%,且检测速度提升了2.7倍,同时降低了安全帽佩戴检测的漏检率以及误检率。

Aiming at the problem that low identification accuracy,high false identification rate and missed identification rate and slow identification speed in helmet wear identification under complex industrial scene,an improved YOLOv3 helmet wear identification algorithm with high recognition accuracy and fast identification speed was proposed.The traditional YOLOv3 trunk network was trimmed and modified to improve the identification speed.The spatial pyramid pooling module was introduced to integrate local features and global features more effectively.In order to improve the fitting effect between the target prediction frame and the real target frame,the loss function was improved to CIoU.Finally,the fourth feature fusion scale was extended for small target identification to improve the accuracy of small target recognition.The results show that the improved YOLOv3 improves the average identification accuracy of helmet wear identification by 2.37% and the identification speed by 2.7 times in complex industrial environment,and at the same time,the missed identification rate and false identification rate of helmet wear identification are reduced.