摘 要:近年来,涉恐涉爆案件频发,为进一步加强对人流密集场所的安保,充分利用新兴的视频分析和人脸识别比对技术有针对性的开展安全防范工作,在目前人脸识别算法所能达到的识别高度前提下,可通过多阈值技术有效提高动态人脸识别情况下公安处警精确度。该方法能有效降低动态人脸识别系统在不同应用环境中的误报率和漏报率,提高动态人脸识别在公安的实战应用效果。
关键词:人脸识别;多阈值;动态监控
中图分类号:TP391.41 文献标识码:A 文章编号:2096-4706(2018)01-0006-04
Actual Combat Appilication for Police with Multi Threshold Based on Dynamic Facial Reconigtion
JIA Yonghong,YAN Zhuanyi,HE Yongkeng,XIAO Boming
(Guangzhou Municipal Public Security Bureau,Science and Technology Communications Department,Guangzhou 510030,China)
Abstract:In recent years,there have been frequent cases of terrorist involvement,in order to further strengthen the security of crowded places.We should make full use of emerging video analysis and face recognition technology to carry out safety precautions against the premise that the face recognition algorithm can achieve high recognition accuracy.Can effectively improve the dynamic face recognition under the condition of public security alarm accuracy by multi threshold technique.This method can effectively reduce the false positive rate and false negative rate of dynamic face recognition system in different application environments,and improve the actual application effect of dynamic face recognition in public security.
Keywords:facial recognize;multi threshold;dynamic monitoring
参考文献:
[1] https://www.nist.gov/programs-projects/facevideo-evaluation-five.
[2] 柴秀娟,山世光,卿来云,等. 基于3D 人脸重建的光照、姿态不变人脸识别 [J]. 软件学报,2006,17(3):525-534.
[3] 张志伟. 可变光照条件下的人脸识别技术研究 [D]. 河北工业大学,2007.
[4] 卢春梅,牛海军,郝琳波. 消除光照和姿态变化影响的人脸识别 [J]. 计算机工程与应用,2008,44(10):166-168.
[5] 刘会东. 姿态变化与困难光照条件下的人脸识别研究 [D].南京师范大学,2013.
[6] 朱进. 人脸识别中光照影响的研究 [D]. 吉林大学,2007.
作者简介:贾永洪,男,广州市公安局科技通信处高级工程师, 广州市“121 人才梯队工程”后备人才。