基于改进BP神经网络的人脸识别算法

Face Recognition Algorithm Based on Improved BP Neural Network

神经网络具有运行计算速度慢,不容易收敛的缺点,文中针对此问题提出了图像的光照校正、图像降维与改进型神经网络相结合的人脸识别算法。运用了图像进行光照校正,人脸图像进行降维及不同的光照条件下的人脸图像运用改进型的BP神经网络对进行识别。讨论了基于网络中的参数数据选择问题,对网络学习速度和Sigmoid函数进行了明显改善。实验结果表明,其识别率有了显著的提高;改进后的BP网络收敛速度在得到相同识别率的效果下显著加快。

As the neural network computing to run slow,not easy to convergence problem,put forward for this image illumination correction,image dimensionality combined with the improved neural network face recognition algorithm.Use the image illumination correction,reduce the dimension of face images and different lighting conditions,the use of human face images improved the BP neural network for recognition.Also discussed the parameters of the network-based data selection problem,learning speed and Sigmoid function was significantly improved.The experimental results show that the recognition rate has improved significantly;improved convergence rate of BP network get the same recognition rate in effect,were significantly accelerated.