视频监控场景下的面部遮挡检测

Face occlusion detection for video surveillance

提出了一种监控场景下的面部遮挡检测方法。基于Ada Boost算法进行人脸验证,通过面部划分,分块分析是否存在遮挡情况。首先判断是否有人进入,在有人进入的情况下进行面部遮挡检测,对眼部区域采用Ada Boost方法及墨镜特征提取方法判断是否遮挡,而对嘴部区域采用高斯肤色模型进行判断。实验结果表明,该方法能实时检测面部遮挡的情况,并达到了较好的效果,适用于银行ATM等监控场景,具有较高的应用价值。

A facial occlusion detection method for surveillance scene is proposed. Face identifying is based on the Ada Boost algorithm. The face is divided into two blocks, and each block is analyzed to determine whether it is occluded or not. The method detects whether someone enters, then determines whether the face is occluded. The Ada Boost algorithm and sunglasses feature extraction are used for eye region, and Gaussian color model is used for mouth region. Experiments indicate the proposed method shows high accuracies on the test videos. It＇s suitable for bank ATM surveillance scene, and has a high application value.