基于CMYK-H-CbCr肤色检测和改进型AdaBoost算法的人脸检测

摘要

针对复杂背景和高分辨率的人脸检测问题,提出一种多颜色空间下的肤色检测和改进型AdaBoost算法结合的人脸检测方法.首先,为了提高检测速度,采用多颜色空间的肤色检测作为预处理,结合CMYK、HSV、YCbCr三种颜色空间下的肤色阈值分割,得到人脸候选区域;其次,为了克服人脸相似区域容易导致的退化现象,将样本和弱分类器阈值的距离结合到权重更新中,提出一种改进型的AdaBoost算法.实验证明,二者结合后的新方法,在保证检测率的同时,大幅降低了计算复杂度和误检率.%Aiming at complex background and high resolution face detection problem, a face detection method combining skin color detection and improved AdaBoost algorithm in multi-color space is proposed.First, in order to improve the detection speed, this paper uses the skin color detection of multi-color space as the pretreatment, and combined with CMYK, HSV, YCbCr three color space under the skin color threshold segmentation to get the face candidate area.Second, in order to overcome the degradation caused by the similarity of the similarity region, this paper combines the distance between the sample and the weak classifier threshold into the weight update, and proposes an improved AdaBoost algorithm.The experimental results show that the new method greatly reduces the computational complexity and false detection rate, while ensuring the detection rate.