一种基于改进样本选取的快速人脸检测方法

A FAST FACE DETECTION METHOD BASED ON IMPROVED SAMPLE SELECTION

AdaBoost算法效果的好坏关键在于前期训练时候的弱分类器的选取,而弱分类器的选取在一定程度上依赖于样本集的选取。因而训练样本集显得十分重要。深入分析了cascade分类器与弱分类器之间的关系,从样本选取角度出发,根据检测率、漏检率与错检率三个指标,改进样本选取,提出一种快速人脸检测方法,该方法分为训练和检测两部分,主要通过对训练样本的比例优化和检测窗口的合并来实现。实验结果表明,该方法检测性能上比传统方法有更好的检测效果。

Whether AdaBoost algorithm runs well or not in effect relies on the weak classifier selection in its early training time,however the selection of weak classifier depends on the selected sample set to a certain extent.Thus the training sample set is very important.In this paper,the relationship between cascade classifier and weak classifier has been analyzed in detail.According to the detection rate,the undetected rate and the false detection rate and proceeding from the sample selection,we improve the method of sample selection and propose a fast face detection method which is divided into training and testing.It mainly optimises the proportion of training samples and merges the detection window to achieve the purpose.Experimental results show that the proposed method has much better detection effects than traditional method in detection performance.