基于Logistic回归和BPNN的二值人脸图像识别

BINARY FACE IMAGE RECOGNITION BASED ON LOGISTIC REGRESSION AND BPNN

针对人脸检测识别中受外在条件影响及低识别率的问题,提出一种基于二值图像的Logistic回归和反向传播神经网络BPNN(back-propagation neural network)的人脸识别方法。该算法将彩色图像被转换成灰度图像。使用低通滤波器去噪,将局部窗口标准偏差和自适应阈值应用于灰度图像,得到高质量的二值去噪图像,从中检测可能的人脸区域。使用最近邻居内插方法将其缩小,与每个缩小大小的图像相对应地创建人脸数据库。使用Logistic回归和BPNN来分类属于每个人的所有图像,并为每一类图像获得一个决策边界。图像尺寸的缩小最大限度地减少了逻辑回归和神经网络训练的计算空间和时间。实验结果表明,在FEI图像数据库上Logistic回归和反向传播神经网络的识别精度高达97.5%,优于其他识别算法的精度。

Face detection and recognition is affected by external conditions has low recognition rate. To solve this problem, we proposed a face recognition method based on binary image Logistic regression and back-propagation neural network (BPNN). The algorithm converted the color images to grayscale images. A low-pass filter was used to remove the noise. The local window standard deviation and adaptive threshold were applied to grayscale images so as to obtain high-quality binary denoised images from which possible face regions were detected. We adopted the nearest neighbor interpolation method to reduce them. A face database was created corresponding to each reduced-size image. We used Logistic regression and BPNN to classify all images belonging to each person, and obtained a decision boundary for each type of image. The reduction in image size minimized the computational space and time of logistic regression and neural network training. Experimental results show that the accuracy of Logistic regression and BPNN is as high as 97.5% in the FEI image database, which is better than the accuracy of other recognition algorithms.