Abstract: A fast face detection algorithm in video integrated with tracking is designed. It uses a cascade-structured face detector trained with AdaBoost to detect face. Integrated with a linear face tracker, searching region and rescaling levels of the detector are reduced, and thus detection speed is accelerated. It also designs a classifier training method integrated with AdaBoost, bootstrapping and weight-trimming, which can be easily used in object detection.

Key words: Face detection; Face tracking; Face recognition

摘要： 提出了一种结合人脸检测和跟踪的快速动态人脸检测算法。在层次型AdaBoost 检测算法的基础上，结合提出的一种基于线性预测的人脸跟踪算法来缩小搜索空间，使得检测速度提高到24ms/帧。还设计了一种结合层次型AdaBoost 算法、负样本反馈方式和利用权值缩减训练时间的检测器训练方法，可以方便地应用于物体检测技术上。

关键词: 人脸检测；人脸跟踪；人脸识别