主动毫米波成像隐私保护算法  

A privacy protection algorithm for active millimeter-wave imaging

作　　者：叶金晶[1,2] 周健[1] 孙谦晨 杨明辉[1] 朱玉琨[1,2] 孙晓玮[1] 

机构地区：[1]中国科学院上海微系统与信息技术研究所中科院太赫兹固态技术重点实验室,上海200050 [2]中国科学院大学,北京100049

出　　处：《红外与毫米波学报》2017年第4期505-512,共8页Journal of Infrared and Millimeter Waves

基　　金：国家自然科学基金(61671439);中国科学院院地合作项目(YDJDBNJ-2012-027);上海市自然科学基金(13ZR1447400)~~

摘　　要：针对主动毫米波成像安检系统的个人隐私保护问题,提出了一种不受身高、站姿影响的毫米波人体图像隐私保护算法.通过预处理将毫米波图像背景噪声弱化,增加人体目标边缘细节.使用二值化和形态学处理进一步区分背景与人体目标,根据身高划分的人体部位比例与像素点遍历求最值法实现隐私部位的初步定位,可以满足标准站姿下不同身高受检人的隐私部位定位需求.算法所包含的定位复查校准模块通过对隐私位置条件的循环判断与调整,解决了非标准站姿下隐私部位的定位.最后,根据已定位的坐标进行模糊化处理,实现保护受检人隐私的功能.算法已应用于实际产品,与常规隐私保护算法相比,可移植性高,运行速度快。A kind of human body image privacy protection algorithm was proposed for active millimeter- wave imaging security systems with the images was not affected by the height and the standing posture of people. The background noise of the millimeter-wave image was decreased and the edge details were increased by pretreatment of the human body image. The morphological processing and binarization were used to distinguish human body image from background. The pixel traversal method and human body proportion determination method were initially used to locate the private parts of human body with standard posture. Subsequently, the calibration algorithm was used to locate the privacy parts of human body with different height and non-standard posture to improve the identity rate of location. Finally, the already positioned privacy parts were blurred to protect the privacy of people under inspection. The algorithm has been applied to the actual imaging products. Compared with the conventional ones, the proposed algorithm has high portability, fast running speed and high positioning accuracy.

关 键 词：毫米波 隐私保护 定位 图像处理 

分 类 号：TN911.73[电子电信—通信与信息系统]