基于YOLO-MCA的X光图像检测算法

X-ray image detection algorithm based on YOLO-MCA

YOLO算法直接用于X光图像检测时存在提取特征不明显问题,特别是违禁物与安全物存在折叠交叉时,容易导致漏检、多检现象。为此本文提出一种YOLO-MCA算法,该算法在YOLOv5基础上,增加了一个多卷积融合坐标注意力机制分支模块,该模块通过多支路连通的方式增大感受野,注重位置信息提取,增强提取有效特征能力,可改善物体折叠交叉导致的漏检、多检问题。在PIDray_OD数据集上的实验结果表明,所提出的YOLO-MCA算法的mAP@0.5∶0.95达到72.9%,比原模型算法的精度更高;FPS达到87,满足实时检测需求。

The YOLO algorithm is directly used for X-ray image detection,and the feature extraction is not obvious,especially when the contraband and the safe object are folded and intersected.It is easy to lead to the problem of missed detection and multiple detection.Therefore,this paper proposes a YOLO-MCA algorithm based on YOLOv5,and it adds a multi-convolution fusion coordinate attention mechanism branch module,which increases the receptive field through multi-branch connectivity,and at the same time it makes the network focus on the extraction of location information and enhances the extraction of more effective features.It effectively solves the problem of missed inspection and multiple inspection caused by the folding and crossing of objects.The experimental results on PIDray_OD dataset show that the mAP@0.5:0.95 of the proposed YOLO-MCA algorithm reaches 72.9%,which is higher than the accuracy of the original model algorithm;and that FPS reaches 87,meeting the needs of real-time detection.