基于特征分析的人脸检测算法的DSP实现  

Implementation of Face Detection Algorithm Based on Feature Analysis in DSP System Board

作　　者：刘纪红[1] 刘祚[1] 李中帆[1] 

机构地区：[1]东北大学信息科学与工程学院,沈阳110819

出　　处：《电子器件》2013年第6期774-778,共5页Chinese Journal of Electron Devices

摘　　要：由于监控视频中人脸影像的多角度性和多尺度性,实时视频中人脸的实时检测成为智能监控系统中的关键技术之一。算法首先根据MATLAB统计好的肤色模型确定候选肤色区域,然后依据人脸的形状特征排除非人脸区域,可以实现所有人脸的精确定位;同时采用区域增长的连通域方法来确定人脸候选块,代替了边缘检测等方法,减少了DSP的运算量。算法在ICETEK-DM6437-B评估板上实现。经过测试,该算法可以很好的满足实时人脸检测的要求。Due to the multi-angle and multi-scale properties of the face image in surveillance video, real-time face detection becomes one of the important technologies in intelligent surveillance system. Firstly,the skin color model, which is analyzed and obtained in MATLAB platform, is used to determine the candidate skin regions. Secondly, according to the shape features of people＇s faces, the non-face areas are excluded, and this can achieve the precise positioning of the faces. In the process of determining the face candidate region, the method of connected domain of the regional growth is used to determine the face candidate block instead of other methods such as edge detection so that it can reduce the amount of computation of the DSP. The algorithm is realized on the ICETEK-DM6437-B system board. The final test shows that the algorithm can meet the requirements of real-time face detection well.

关 键 词：模式识别 人脸检测 特征分析 DSP 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]