基于驾校考勤机的人脸检测应用研究

Application Research on Face Detection Based on Driving School Attendance Machine

针对驾校现有指纹考勤系统存在的学时造假及部分指纹检测失效等问题,采用了一种椭圆高斯双肤色模型与Adaboost快速检测算法相结合的人脸检测算法。首先,应用HSV-自适应Retinex算法对待检测图像进行光照增强处理。其次,在聚类性较好的YCgCr颜色空间对光照增强后的图片进行双肤色建模,粗检测出人脸区域。最后,利用形态学方法去除孤立肤色点以缩减检测范围,并运用Adaboost快速检测算法实现人脸的精确定位。实验数据显示,该方法在检测率、检测时间等方面有所提高。实验结果表明,改进后的人脸检测算法可实现光照不均以及不同年龄下的人脸检测,能够快速且较为准确的检测人脸。

Due to the driving school existing fingerprint attendance system problems such as hours fraud and part of the fingerprint detection failure, a algorithm for face detection combining elliptical Gaussian double skin color model and rapid detection of Adaboost algorithm was put forward. First of all, a method based on HSV-adaptive Retinex algo- rithm was proposed to illumination compensation in order to make the photos in uniform illumination environment. The method can solve the inadequate light of vehicle cab. What＇s more, due to the clustering is excellent in YCgCr color space. A double skin model based on Elliptical and Gaussian model was put forward to deal with the picture which have been illumination compensation. Then, the face regions were crudely detected. Finally, using morphological methods to remove isolated skin spot, in this way can reduce the detection range. And using Adaboost algorithm, the accurate face detection was acquired. Experimental data show that this method has increased in terms of the detection rate, detection time. Experimental results indicates that the improved face detection algorithm which can realize the face detection in uniform illumination and different ages, able to quickly and more accurately detect human faces.