摘要
针对安检通道中人工检测违禁物品导致效率低的问题,提出一种基于全卷积网络改进的X光图像违禁物品检测方法。采用单阶段检测算法提高计算速度和简化模型,实现无锚框情况下的逐像素检测;构建双向特征金字塔进行多尺度融合,一定程度上解决物品重叠的问题并降低漏检率;改进损失函数实现高效的模型训练,减少内存占用和低质量的计算。实验结果表明,所提算法能够在民航安检应用中实现准确、高效的智能X光图像中违禁物品的检测。
To solve the problem of low efficiency of manual detection of prohibited item in security inspection channels,an improved X-ray image detector based on fully convolutional networks was proposed.The one-stage object detector was adopted to improve the calculation speed and simplify the model,and the per-pixel detection was realized in the case of anchor free.To improve the detection accuracy,the bi-directional feature pyramid was constructed to perform feature fusion at different levels,which alleviated the problem of overlapping items and reduced the rate of missing detection.Loss function was improved to achieve efficient model training,reducing memory saving and low quality calculations.Experimental results show that the proposed method can achieve accurate and efficient X-ray detection of prohibited items in civil aviation security.
作者
李舒婷
姜永峰
张良
LI Shu-ting;JIANG Yong-feng;ZHANG Liang(College of Electronic Information and Automation,Civil Aviation University of China,Tianjin 300300,China;Intelligence Information Detachment,Wenzhou Public Security Bureau,Wenzhou 325000,China)
出处
《计算机工程与设计》
北大核心
2021年第11期3188-3195,共8页
Computer Engineering and Design
基金
国家自然科学基金项目(61179045)。
关键词
安检判图
全卷积网络
X光图像
违禁物品检测
双向特征金字塔
civil aviation security inspection
fully convolutional networks
X-ray images
prohibited items detection
bi-directional feature pyramid network
