一种改进的Adaboost算法的人脸检测分类器

针对人脸检测中单一分类器检测非常耗时,而且没有考虑到检测后提取标准人脸对后续工作(如识别)等问题,提出了运用Adaboost算法,采用正面人脸和人眼两个特征作为分类器,首先用Cascade算法筛选出输入图像中最有可能是人脸的区域,然后根据以正面人脸和人眼为特征的分类器检测人脸区域,并根据检测出来的人眼距离和角度对人脸进行精确的定位.同时针对实际的图像中人脸常常存在一定倾斜,从而影响后续的人脸的识别率这一问题,采取了根据人眼的位置对倾斜图像进行修正.实验结果表明改进的算法能够在保持一定运算速度的基础上取得准确的人脸,具有较好的实用性.

Research on Face Detection Classifier Based on An Improved Adaboost Algorithm

Aimed at the questions in face recognition by using single classifier,such as the time consuming of single classifier detection,and the significant of face image extracting after detection etc,this paper proposes a method based on Adaboost algorithm to handle it.The classifiers based on the facade and the eyes.First,we screen out the most probably facial area by the Cascade algorithm,then,we detect the facial area with the classifiers,finally,we will perform the precise positioning to the face image with th...