X射线图像违禁品自动检测算法研究

Research on Automatic Detection Algorithm for Contraband in X-ray Images

为了提高城市轨道交通X射线安检违禁品自动识别的有效性、提升安检效率,设计了一种带有注意力机制的YOLOv5m算法模型。该算法模型引入置换注意力(Shuffle Attention,SA)模块,采用通道随机混合操作,分块并行使用空间和通道两类注意力机制,使两类注意力机制高效结合。通过实验验证,改进后的算法的mAP50达到了0.718,相较于基础的YOLOv5m算法,mAP50:95和mAP50分别提升了0.8个百分点和1.2个百分点,结果表明改进后的算法能显著提高检测识别精度。

In order to improve the effectiveness of automatic detection of contraband in X-ray security checks of urban rail transit and improve security efficiency,a YOLOv5m algorithm model with attention mechanism was designed.This model introduces the Shuffle Attention(SA)module.The module uses channel random mixing operations,and parallelizes the use of spatial and channel attention mechanisms in blocks,enabling efficient integration of the two types of attention mechanisms.Through experimental verification,the mAP50 of the improved algorithm reaches 0.718;mAP50:95 and mAP50 are respectively 0.8 percentage points and 1.2 percentage points higher than the basic YOLOv5m algorithm.The results show that the improved algorithm can significantly improve detection accuracy.