摘要
使用Adaboost算法实现人脸检测会出现一定的误检率。针对这一问题,设计了一种在误检情况下的识别系统,对待识别图像先使用grabcut前景检测算法进行前背景分割,在一定程度上消除环境因素的影响,然后对分割结果进行人脸检测和识别。该系统检测部分使用haar级联分类器,识别部分使用特征脸算法。实验结果表明,结合grabcut和Adaboost算法系统在识别率和检测率方面均有一定提高,且识别速度较快。
For using Adaboost algorithm to recognize human-face would occur a corresponding false alarm rate.This essay has designed and implemented a face recognition system using grabcut algorithm for image to be identified to segment the foreground and background at first when false alarm rate occurred,then using haar cascade classifier and Eigenface algorithm to recognize human-face in image.Experimental results show that in the human-face recognition system using grabcut and Adaboost algorithm relatively improved recognition rate and detection accuracy,the recognition speed could basically satisfy the practical application.
出处
《软件导刊》
2017年第12期99-101,105,共4页
Software Guide
基金
国家自然科学基金项目(61565014)