【摘要】 针对人脸部分被遮挡后检测率较低的问题,提出一种基于运动信息的快速局部遮挡人脸检测方法。该方法在训练阶段采用快速抽取关键样本算法减少训练时间,并初始化训练样本权值,在检测阶段首先分析视频帧间的运动信息估计人脸的大概位置,然后通过瀑布型快速人脸检测器进行进一步精确定位。实验结果表明,该方法在人脸局部遮挡情况下,相对于传统的Adaboost算法检测率有了明显提高,同时具有较快的检测速度。更多还原

【Abstract】 Aiming at the low detection rate on the partial occluded face,a fast partial occluded face detection method based on motion information is proposed in this paper.In the stage of training,a rapid extraction representative sample algorithm is used to reduce training time and initialize the weight of representative samples.In the stage of detecting,the motion information in the video frames is first analysed for estimating probable location of the face.And then the face is further located accurately by cascadic fast face detector.Compared with traditional Adboost algorithm,the experimental results demonstrate that the proposed approach obviously improves the occluded face detection rate,and also has fast detection speed as well.更多还原