基于BP神经网络的人脸检测AdaBoost算法

AdaBoost Algorithm for Face Detection Based on BP Neural Network

人脸检测在日常生产和应用非常重要;现有的人脸检测算法存在精确度不高、运算复杂等问题,为了解决目前存在的这些问题,提出了一种基于BP神经网络的AdaBoost人脸检测算法;首先,使用BP神经网络代替YCbCr高斯模型建立肤色模型;同时,针对AdaBoost算法提出了一种新的权值更新方法;在权值更新中引入阈值与样本之间的距离;另外权重有一个边界值;最后,利用BP神经网络提取图像中的肤色候选区域,并采用改进的AdaBoost算法对图像中的人脸进行精确检测;实验结果表明,利用BP神经网络和改进的AdaBoost算法的新的解决方案比现有的方法具有更高的精度,算法精确度达94%左右。

Face detection is very important in daily production and application.Existing face detection algorithms have low accuracy and complicated calculations.In order to solve these problems,an AdaBoost face detection algorithm based on BP neural network is proposed.First,a BP neural network was used instead of the YCbCr Gaussian model to build a skin color model.At the same time,a new weight update method is proposed for the AdaBoost algorithm.The distance between the threshold and the sample is introduced in the weight update.In addition,the weights have a boundary value.Finally,the BP neural network is used to extract the skin color candidate regions,and the improved AdaBoost algorithm is used to accurately detect the faces in the images.Experimental results show that the new solution using BP neural network and improved AdaBoost algorithm has higher accuracy than the existing methods,and the algorithm accuracy is about 94%.