摘要
研究了灰度值、中值滤波的图像预处理方法和Haar特征提取思想计算多尺度下相同特征。本文基于Adaboost算法针对同一个训练集训练不同的分类器,并将弱分类器进行集合,构成一个更强的最终分类器,实现了脸谱识别系统。通过验证脸谱识别系统,实现了对视频流中脸谱的准确定位,达到了无拖影、噪声少及识别准确的预期。
In the paper, the image preprocessing method of gray value and median filter are studied, then the same feature of Haar feature extraction are researched.The different classifier based on Adaboost algorithm for the same training set is designed.The weak classifiers are set to form a stronger final classifier,and finally the face recognition is achieved.By verifying the faeebook recognition system,the ac- curate positioning of the face in the video stream has achieved no smear,less noise,and accurate prediction.
出处
《单片机与嵌入式系统应用》
2017年第8期29-32,共4页
Microcontrollers & Embedded Systems
基金
山东省自然科学基金资助(ZR2015DM013)
国家自然科学基金(41572244)