复杂光照条件下基于深度学习的人脸识别方法

Deep Learning-Based Face Recognition Method Under Complex Illumination Conditions

由于传统方法在复杂光照条件下人脸识别的识别时间比较长、误识率较高,提出复杂光照条件下基于深度学习的人脸识别方法。首先,利用相机与灯光组合拍摄不同光照条件下人脸图像,建立复杂光照条件下人脸数据集。其次,对人脸图像进行对数转换、直方图均衡化以及自适应滤波处理。最后,建立深度学习网络模型,利用模型提取和识别人脸特征。分析实验可知,在复杂光照条件下,设计方法人脸识别时间在1 s以内,误识率在1%以内。

Due to the long recognition time and high error rate of traditional methods for face recognition under complex lighting conditions,a face recognition method based on deep learning under complex lighting conditions is proposed.Use a camera and light combination to capture face images under different lighting conditions,and establish a face dataset under complex lighting conditions.Face image logarithmic conversion,histogram equalization,and adaptive filtering processing.Establish a deep learning network model,and use the model to extract and recognize facial features.According to the analysis and experiment,under complex lighting conditions,the face recognition time of the design method is within 1 s,and the error recognition rate is within 1%.