中文摘要：

在复杂背景下的人脸检测技术是当今智能视觉技术中的一项难题。为了提高人脸检测的精度和实时性,降低误检率,基于HSV模型和模糊级联分类器对复杂背景中的人脸检测技术进行研究。首先在HSV模型中对图像进行选择性光线补偿,然后对图像在HSV模型中进行分割,接着用图形学的方法去噪,再将连通的肤色区域构建肤色团块,并且利用人脸的脸部比例特征来剔除不相符的人脸团块,最后利用模糊级联分类器对肤色团块检测人脸。该算法的误检率和漏检率分别为0.1%和5.9%,检测的准确率可以达到94.1%,并且有效提高了检测速度,具有一定的实用价值。实验结果表明,基于HSV肤色检测和模糊级联分类器的算法能更好地处理人脸在较差光线和有阴影干扰的环境下的检测。

英文摘要：

It is one of a present intelligent vision technology problems for face detection technology in a complicated background. In order to improve face detection precision and timeliness and reduce its false detection rate,based on HSV model and fuzzy cascade classifier,the paper studies the face detection technology in a complicated background. Firstly selective illumination compensation is made for the pictures in the HSV model. Secondly pictures are segmented in the HSV model. Thirdly noise is eliminated by graphics method. Fourthly skin color blocks are built from continuous skin color areas and unsuitable face blocks are removed according to face scale characteristics. Finally the fuzzy cascade classifier detects face from skin color blocks. The error rate and miss rate for detection of the algorithm is 0. 1% and 5. 9% respectively. Its correct detection rate reaches 94. 1% while it effectively improves the detection speed. So it is of practical value. Experiment result illustrates that the algorithm based on HSV skin color detection and fuzzy cascade classifier can better handle face detection in dim-light or shadow-interfered environments.