基于高光谱的低对比度伤员目标搜寻技术研究  

Research on Low-Contrast Wounded Human Target Search Technology Based on Hyperspectral

作　　者：朱明明 尚丹婷 曹育森 雷涛 夏娟娟 李钊 景裕 张林媛 王健琪 路国华 ZHU Mingming;Shang Danting;CAO Yusen;LEI Tao;XIA Juanjuan;LI Zhao;JING Yu;ZHANG Linyuan;WANG Jianqi;LU Guohua(Shool of Military Biomedical Engineering,Air Force Medical University,Xi’an Shaanxi 710032,China;Drug and Instrument Supervisory&Test Station,Xining Joint Logistics Support Center,Lanzhou Gansu 730050,China;Department of Radiology,Tangdu Hospital,Air Force Medical University,Xi’an Shaanxi 710038,China)

机构地区：[1]空军军医大学军事生物医学工程学系,陕西西安710032 [2]西宁联勤保障中心药品仪器监督检验站,甘肃兰州730050 [3]空军军医大学唐都医院放射科,陕西西安710038

出　　处：《中国医疗设备》2021年第6期5-8,25,共5页China Medical Devices

基　　金：国家自然科学基金面上项目(61271102);军队后勤科研计划重大子项(AS314Z004-01);陕西省重点产业创新链(2021ZDLGY09-07)。

摘　　要：目的为解决低对比度环境中伤员难以被快速发现的问题,本文提出了一种基于无人机搭载高光谱的伤员识别方案。方法使用M600六旋翼无人机搭载HY-1010高光谱相机采集地面伪装伤员遥感图像,仿真血喷涂迷彩服模拟低对比度环境中伤员,高光谱遥感图像经过预处理后,提取目标波谱和背景波谱,采用混合调谐匹配滤波、最小能量约束(Constrained Energy Minimization,CEM)、自适应一致估计(Adaptive Coherence Estimator,ACE)三种常见的高光谱地物识别方法进行伤员识别。结果三种识别方法均可识别出模拟的6个伤员,从像素级统计分析得出ACE方法误识别率最低53.2%,CEM识别正确率最高50%。结论该实验采用高光谱识别技术可识别出草地环境中迷彩服上的血迹,为低对比度环境下伤员的识别提供新的思路。Objective To improve the rate of the wounded identification and solve the problem of finding the wounded in similar colour clothes quickly under low-contrast background environment,a new identification scheme based on hyperspectral of uncrewed aerial vehicle is proposed.Methods The remote sensing images of the camouflaged wounded on the ground were collected by the M600 six rotor UAV equipped with HY-1010 hyperspectral camera,and the simulated blood spraying camouflage clothes were used to simulate the wounded under low-contrast background.After the hyperspectral remote sensing images were preprocessed,the target spectrum and background spectrum were extracted,and the three common hyperspectral ground object recognition methods of mixed tuned matched filtering,constrained energy minimization(CEM)and adaptive coherence estimator(ACE)were used to identify the wounded.Results The three recognition methods can identify six simulated wounded.From the statistical analysis of pixel level,the false recognition rate of ACE method was the lowest 53.2%,and the correct recognition rate of CEM method was the highest 50%.Conclusion This experiment uses hyperspectral recognition technology to identify the bloodstain on the camouflage clothes in the grassland environment,which provides a new idea for the identification of the wounded under low-contrast background environment.

关 键 词：低对比度 高光谱 识别 伤员 

分 类 号：R197.39[医药卫生—卫生事业管理]