CAJ | 학술논문

根据可见光图像与红外图像的信息互补性,分析在决策层融合识别中的归一化法和融合算法,提出一种基于统计的Fisher投影融合法,利用Fisher线性判别准则在二维分数空间寻找最优投影方向,使不同类样本投影后能最佳分离.在多光谱人脸融合识别中的应用结果表明,与其他融合算法相比,该算法具有更好的识别效果.
근거가견광도상여홍외도상적신식호보성,분석재결책층융합식별중적귀일화법화융합산법,제출일충기우통계적Fisher투영융합법,이용Fisher선성판별준칙재이유분수공간심조최우투영방향,사불동류양본투영후능최가분리.재다광보인검융합식별중적응용결과표명,여기타융합산법상비,해산법구유경호적식별효과.
According to the information complementarity for visible image and infrared image, this paper analyzes normalization method and fusion algorithm for fusion recognition at decision level, proposes a Fisher projection fusion algorithm based on statistics. This algorithm employs Fisher linear discriminate criterion to find out the optimal projection in score space that can best separate projected samples from different classes. Application of multi-spectral face fusion recognition results show that this algorithm has higher recognition effect compared with other fusion algorithm.