基于运动区域Viola-Jones算法的视频人脸检测  

Face detection in video motion region based on Viola-Jones algorithm

作　　者：彭明莎 刘翠响[1] 

机构地区：[1]河北工业大学,天津300401

出　　处：《电子设计工程》2015年第21期15-17,共3页Electronic Design Engineering

基　　金：校博士科研基金项目(2009032012)

摘　　要：为了进一步提高人脸检测的准确率,本文采用一种基于运动区域的检测方法,该方法主要通过采用帧差法和自适应的滑动平均法相结合的算法,建立提取运动区域的模型,然后通过基于Boxfilter算法的Viola-Jones进行快速人脸检测。自适应的滑动平均法是通过检测光照突变和像素点来实现阈值和更新率的自适应更新,该法提取的前景与帧差法提取的前景进行"与运算"。实验结果表明,该方法可以有效地提取运动区域,从而使人脸检测性能得到进一步的提高,具有一定的应用价值。In order to improve the accuracy of face detection, this paper adopts a detection method in motion region. The method to extract moving regions model is established, through a algorithm of combining frame difference method and adaptive moving average method. Then faces are detected through Viola-Jones algorithm based on Boxfilter . The moving average method adoptes threshold and update rate by detecting the light adaptive mutation and the pixel. ＆#39;and operators’ is on the extracted foreground and that is extracted by frame difference method .The experimental results show that, this method can effectively extract moving regions. So that the face detection performance can be further improved, and it has certain application value.

关 键 词：帧差法 滑动平均法 Viola-Jones算法 Boxfilter算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]