一种基于卷积神经网络的违禁品探测系统及部署方法

A Contraband Detection System and Its Deployment Method Based on Convolutional Neural Network

文章设计了一种X射线安检图像的违禁品自动识别系统。以经典的ResNet卷积神经网络为特征提取骨干网络,使用特征金字塔网络进行违禁品目标探测;把图像分成笔记本区域和非笔记本区域,分别用不同的网络模型进行探测,取得了出色的探测效果。并使用网络服务器的模式部署上述算法探测系统,使得产生图像的安检机系统与算法服务器既保持彼此独立又能进行稳定、实时的通讯,完成探测。

This paper proposes a contraband detection system for X-ray security inspection machines. Based on the classic ResNet convolutional neural networks as backbone networks, we detect contrabands with Feature Pyramid Networks. A better performance is achieved by segmenting an image as notebook area and no-notebook area which are processed separately with different detection models. And the models are deployed in a network server connected with the X-ray security inspection machine which sends X-ray images to the server. Both sides communicate with each other stably in a real time.