中文摘要：

针对WMIL在光照突变或者全部遮挡的的情况下会出现跟踪失败以及在跟踪错误情况下无法自动恢复跟踪的问题,提出了一种基于改进WMIL算法和AdaBoost的实时人脸检测和跟踪算法。利用AdaBoost的方法检测人脸信息,然后在改进WMIL算法的框架下,对人脸图像进行多尺度表示,采用压缩感知的方法来提取样本特征。最后,利用改进WMIL算法建立分类器对人脸进行跟踪,自适应调整跟踪窗口的大小,并实时更新。实验结果表明,改善了WMIL存在的不足,有效解决了在人脸外观变化,姿态改变、快速运动等情况下,能稳定准确地实现目标的实时跟踪。

英文摘要：

Under the condition of the light mutation or the total occlusion of the object,the tracking will fail and cannot be automatically recovered by using weighted multiple instanced learning（ WMIL）. Aiming at these problems,a new real-time face detection and tracking algorithm based on the improved WMIL and Ada Boost was presented. The algorithm firstly used Ada Boost to obtain the face detection information,adopted the multi scale image representation under the WMIL framework and extracted the sample features by the compression perception method.Finally,the model classifier was established by the improved WMIL algorithm to track target face. The tracking window scale was adaptively adjusted and updated in real-time. The experiment results show that the algorithm overcomes the traditional WMIL deficiencies and can stably and accurately track the face target in real-time under the situations of the appearance changes,the target posture changes and fast moving,etc.