基于嵌入式Linux的复杂光照人脸实时检测研究  

Research on real-time face detection of complex illumination based on embedded Linux

作　　者：黄世震[1,2] 黄志勇[1,2] 

机构地区：[1]福州大学物理与信息工程学院,福建福州350116 [2]福建省微电子集成电路重点实验室,福建福州350002

出　　处：《微型机与应用》2016年第12期46-48,51,共4页Microcomputer & Its Applications

摘　　要：为了满足嵌入式设备在复杂光照条件下的人脸检测需求,提出了一种基于光照不变特征的人脸检测方法。该方法先基于Retinex理论提取光照不变分量,再以MB-LBP特征进行Ada Boost人脸检测。结合该方法,实现了一种基于嵌入式Linux的人脸实时检测系统。实验与实际使用结果表明,该人脸检测系统具有较高的人脸检出率,且满足实时性要求。In order to meet the demand of the embedded device in the face detection under complex illumination,a face detection method based on the illumination invariant features is proposed. The method is based on the Retinex theory to extract the illumination invariant component,and then Ada Boost face detection based on Multi-scale Block Local Binary Patterns（ MB-LBP） features is executed. According to this method,a real-time face detection system based on embedded Linux is realized. Experimental and practical using results show that the design of the face detection system has a high rate of face detection,and meets the real-time requirements.

关 键 词：人脸检测 嵌入式 ARM RETINEX MB-LBP ADA BOOST 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]