基于Adaboost的红外视频图像疲劳检测算法

针对以往疲劳检测算法普遍存在的受光照条件影响大、检测测速度慢以及可靠性差等问题,本文提出了一种基于Adaboost的疲劳表情快速检测算法。本文算法在不同环境光照的情况下,利用红外光源照明采集获得大量人脸红外图像样本。经过人脸检测定位以后,将人脸区域中眼睛、嘴巴这两个表情信息最集中的关键部位分割出来,用PCA方法分别提取两个子图块的形变特征,分别输入Adaboost训练得到两个分类器。检测时,待检测图像眼、嘴的特征分别通过相应分类器进行判别,将两个分类器的输出进行或运算得到最终的检测结果。该方法正确率高,速度快,具有很好的泛化能力和较强的鲁棒性,能够满足实时应用要求。

An Algorithm of Fatigue Detection by Infrared Images Based on Adaboost

For the problem of the influence of the illumination and speed and reliability of the previous fatigue detection algorithms,this paper proposes a fast algorithm of fatigue detection based on Adaboost.A great lot of infrared face images have been gained in variable environmental illumination using infrared.After the face detection,the features of eyes’and mouths’shapes are extracted from these images using the PCA method.The Adaboost training procedure trains two classifiers to detect fatigue expression using these features.In the detecting procedure,the features of eyes and mouths to be detected are classified by the corresponding classifiers.Then doing the OR operation between the outputs of the classifiers gives the final result.This algorithm has not only high correct rate and fast speed but also a powerful ability to generally use robustness.And the response time of this algorithm satisfies the realtime requirements.