摘要
近些年来,各种人脸检测技术层出不穷。其中,对于人们所关心的检测率和检测时间,基于肤色的人脸检测算法具有较大的优势,而AdaBoot算法则在检测率和误检率方面表现良好。论文立足于这两种算法的基础上,提出了一种运用肤色分割先将疑似人脸区域从背景图片中提取出来再交给AdaBoot分类器来进行人脸检测的方法,期望将肤色人脸检测算法与AdaBoot检测算法进行优势互补,产生一种算法,使其在拥有较好的检测率的同时具有较快的检测时间和较低的误检率。通过实验,这种肤色分割预处理的方法确实可以综合两种算法的优势,具有较短的检测时间和良好的检测率和误检率。
In recent years,various face detection technologies have emerged endlessly.Among them,the face detection algorithm based on skin color has great advantages for the detection rate and detection time that people care about,while the AdaBoot algorithm performs well in detection rate and false detection rate.Based on these two algorithms,the paper proposes a method of using skin color segmentation to first extract the suspected face area from the background image and then hand it to the AdaBoot classifier for face detection.The algorithm complements the advantages of the AdaBoot detection algorithm to produce an algorithm that has a faster detection time and a lower false detection rate while having a better detection rate.Through experiments,this method of skin color segmentation preprocessing can indeed combine the advantages of the two algorithms,with shorter detection time and good detection rate and false detection rate.
作者
邓锋
胡志勇
肖伟明
钟卫为
DENG Feng;HU Zhiyong;XIAO Weiming;ZHONG Weiwei(Wuhan Research Institute of Posts and Telecommunications,Wuhan 430074;Wuhan Hongxin Technology Service Company with Limited Liability,Wuhan 430074)
出处
《计算机与数字工程》
2020年第4期956-959,共4页
Computer & Digital Engineering
