多特征融合的近红外与可见光异质人脸识别

Near infrared and visible heterogeneous face recognition based on multi feature fusion

考虑到传统识别方法忽略了异质人脸图像的特征融合,导致其识别性能较差,利用多特征融合算法提出了近红外与可见光异质人脸识别方法研究。在人脸图像区域中计算了梯度方向均值,定义了异质人脸特征点,引入代价函数进行迭代计算,提取出异质人脸特征点,通过Gabor函数变换,根据人脸识别特征融合的标准,完成异质人脸图像的多特征融合,结合异质人脸识别原理设计,实现了异质人脸的识别。实验结果表明,多特征融合的识别方法的特征识别重复率超过了90%,人脸图像平面颠簸程度在4%左右,可以使得人脸图像平面变得更加平滑。

Considering that the traditional recognition methods ignore the feature fusion of heterogeneous face images,resulting in poor recognition performance,a near-infrared and visible heterogeneous face recognition method is proposed by using multi feature fusion algorithm.The mean value of gradient direction is calculated in the face image area,the heterogeneous face feature points are defined,the cost function is introduced for iterative calculation,and the heterogeneous face feature points are extracted.Through Gabor function transformation,according to the standard of face recognition feature fusion,the multi feature fusion of heterogeneous face images is completed.Combined with the principle of heterogeneous face recognition,the heterogeneous face recognition is realized.The experimental results show that the feature recognition repetition rate of the multi feature fusion recognition method is more than 90%,and the bumpiness of the face image plane is about 4%,which can make the face image plane more smooth.