基于注意力机制的毫米波图像隐匿目标自主识别方法  

Autonomous Recognition Method for Hidden Targets in Millimeter Wave Images Based on Attention Mechanism

作　　者：裴轩 姜济群 史运锋[1] 于海辉[1] 李明[1] PEI Xuan;JIANG Ji-Qun;SHI Yun-Feng;YU Hai-Hui;LI Ming(The First Research Institute of the Ministry of Public Security of P.R.C.,Beijing100048,China)

机构地区：[1]公安部第一研究所,北京100048

出　　处：《机电产品开发与创新》2024年第2期36-39,共4页Development & Innovation of Machinery & Electrical Products

摘　　要：在人体安检中,毫米波成像是一种对人体安全的检测方法。目前,使用平面毫米波扫描人体图像中存在被检测物体较小、边缘模糊等问题,导致传统的目标检测方法实现高准确率的检测变得困难。本文提出应用注意力机制及图像增强方法,提高目标检测方法对人体毫米波图像隐匿物体检测的精确度和召回率。研究结果表明,注意力机制与图像增强方法可提高毫米波图像目标检测的准确率和召回率。In human security inspection,millimeter wave imaging is a safe detection method for human body.At present,there are problems such as small objects and blurred edges in the human body images scanned by planar millimeter waves,which makes it difficult for traditional target detection methods to achieve high-accuracy detection.This study proposes the application of attention mechanism and image enhancement method to improve the precision and recall of target detection method for hidden object detection in human millimeter wave images.The research results show that the attention mechanism and enhancement methods can both improve the precision and the recall in millimeter waves images.

关 键 词：毫米波图像 注意力机制 小目标识别 隐匿物体检测 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]