基于YOLOv3的X光图像违禁品目标检测

Detection of Prohibited Objects in X-ray Images Based on YOLOv3

基于人类视觉的X光图像违禁品检测往往受限于工作强度和复杂人流环境,给安检工作带来巨大挑战。使用人工智能方法对违禁品进行自动检测与判别,对辅助安检工作具有重要的现实意义。提出基于YOLOv3的X光图像违禁品目标检测模型,在传统YOLOv3的基础上增加了一个检测尺度,实现实时的X光图像违禁品的自动判别,即不仅能够辨别出违禁品的种类,还能对违禁品在图像中所处的位置进行标定。实验结果表明,改进后的YOLOv3在Precision、Recall、mAP和F1四个模型评价指标上均取得提高,其中mAP值达到96.2%,对于安检X光图像违禁品目标具有良好的检测效果。

The detection of prohibited items in X-ray images based on human vision is often limited by work intensity and complex human flow environments,which brings huge challenges to security inspection work.The use of artificial intelligence methods for automatic detection and discrimination of prohibited items has important practical significance in assisting security inspection work.This paper proposes a target detection model for prohibited items in X-ray images based on YOLOv3,which adds a detection scale to the traditional YOLOv3 to achieve real-time automatic identification of prohibited items in X-ray images.This model not only identifies the types of prohibited items,but also calibrates the position of prohibited items in the image.The experimental results show that the improved YOLOv3 has achieved improvements in the evaluation indicators of Precision,Recall,mAP,and F1 models,with a mAp value of 96.2%.It has good detection performance for prohibited object targets in X-ray images.