Abstract: This paper proposes a novel method for people counting based on face detection. This method uses AdaBoost algorithm to detect face, count people entering the classroom according to the locations and sizes of the faces appeared in the image. It also detects eyes and uses skin filter to raise the accuracy of face detection. Meanwhile, it uses jumping frame to avoid the false statistic caused by different moving speed of different person. This method is used in a real time face recognition system to count people. Experimental result demonstrates the effectiveness and efficiency of the proposed approach to people counting.

Key words: AdaBoost algorithm, Haar-like feature, face detection, eye detection, skin detection

摘要： 提出一种基于人脸检测的人数统计方法。运用AdaBoost算法进行人脸检测，根据人脸在图像中出现的位置及大小等几何信息统计进入教室的人数。在系统实现中，通过人眼及肤色特征检测提高人脸检测精度，同时利用跳帧检测方法，排除由人员的不同行走速度造成的误差统计。实验结果证明，该方法具有较快的检测及统计速度。

关键词: AdaBoost算法, 类Haar特征, 人脸检测, 人眼检测, 肤色检测

CLC Number: 

• TP391