危险品实时检测定位的Faster RCNN算法研究  

Research on Faster RCNN Algorithm for Real-Time Detection and Location of Dangerous Goods

作　　者：邱钊鹏[1] 朱运利[1] 刘玉娟[1] 林梦圆[1] QIU Zhaopeng;ZHU Yunli;LIU Yujuan;LIN Mengyuan(School of Mechanical and Electrical Engineering,Beijing Electronic Technology Vocational College,Beijing 100176,China)

机构地区：[1]北京电子科技职业学院机电工程学院,北京100176

出　　处：《电子器件》2020年第2期386-390,共5页Chinese Journal of Electron Devices

摘　　要：为了提升安检过程的检测效率以及检测精度,基于Faster-RCNN检测算法,结合VGG-16理论,构建了一种能够实时检测危险品的检测器,结合实际安检采集的影像数据样本,通过深度学习网络对图像数据进行训练,进行数据验证分析,结果表明该检测器具有较高的验证精度,检测算法的精度以及检测效率均高于传统的检测算法,能够较为精准的定位危险品,本文给出的方法为实际安检过程提供了理论支撑及借鉴。In order to improve the detection efficiency and detection accuracy of the security inspection process,based on the Faster-RCNN detection algorithm and VGG-16 theory,a detector capable of detecting dangerous goods in real time is constructed,combined with the image data samples collected by the actual security check,and train the image data through the deep learning network to perform data verification and analysis. The results show that the detector has higher verification accuracy,the accuracy and detection efficiency of the detection algorithm are higher than the traditional detection algorithm,and the dangerous goods can be positioned more accurately. The method provides theoretical support and reference for the actual security inspection process.

关 键 词：卷积特征 深度神经网络 影像识别 危险品检测 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]