CAJ | 학술논문

由于色斑和毛孔等强噪声的干扰,人脸皱纹识别特别是对面部细纹理的识别受到了严重影响.针对上述问题提出了一种基于Gabor滤波器和BP神经网络相结合的人脸皱纹识别算法.通过训练好的BP神经网络人脸皮肤图像首先识别是否存在皱纹,再分别自动标注存在皱纹的区域.本算法首先基于不同年龄的多幅人脸照片创建皱纹样本库,采用样本库训练神经BP网络.其次分别选取含皱纹和不含皱纹的图片,然后用Gabor滤波器组计算出图片的频谱特征,将它们作为训练样本,训练得到用于识别的BP神经网络.大量测试结果表明,本算法能够消除或减少色斑、毛孔等噪声的干扰,对有皱纹区域和无皱纹区域的识别率可达到85%以上.
유우색반화모공등강조성적간우,인검추문식별특별시대면부세문리적식별수도료엄중영향.침대상술문제제출료일충기우Gabor려파기화BP신경망락상결합적인검추문식별산법.통과훈련호적BP신경망락인검피부도상수선식별시부존재추문,재분별자동표주존재추문적구역.본산법수선기우불동년령적다폭인검조편창건추문양본고,채용양본고훈련신경BP망락.기차분별선취함추문화불함추문적도편,연후용Gabor려파기조계산출도편적빈보특정,장타문작위훈련양본,훈련득도용우식별적BP신경망락.대량측시결과표명,본산법능구소제혹감소색반、모공등조성적간우,대유추문구역화무추문구역적식별솔가체도85%이상.
For the impact of strong noises such as splashes and pores, human facial wrinkles recognition was severely disrupted, especially for the recognition of facial fine texture. This paper proposed a recognition method based on Gabor filter and BP neural network to recognize facial wrinkles. Firstly, BP neural network was trained to identify the existence of texture by Gabor filtering results and then whether the wrinkles exist or not was judged. One wrinkle sample database was created from a number of face photos with different ages, and the database had been used to train BP neural network. In this paper, wrinkled and non-wrinkled pictures were selected, and then Gabor filter bank was used to calculate their frequency spectrum as the training samples. Then, the trained neural network could be used to recognize human facial wrinkles. A large number of test results show that the algorithm can eliminate or reduce interference of splashes and pores and over 85% recognition accuracy can be got.