摘要
针对非均匀光照和局部遮挡因素而干扰维吾尔族人脸识别效果,影响了维吾尔族人脸的特征提取效率和维吾尔族人脸识别正确率的问题,提出了基于全局特征和局部特征联合稀疏编码的维吾尔族人脸图像识别算法。首先,进行全局和局部的维吾尔族人脸特征图像的提取进而组成多簇字典;然后维吾尔族人脸图像的特征向量在其固有的特征字典上构建对应的联合编码,然后把不同特征的二维编码系数的方差进行最小化模式,以便使维吾尔族人脸图像的全局特征和局部特征的表现出固有的相似性,并且通过构造不一样的维吾尔族人脸图像特征编码系数和权不同距离来表征其图像特征出贡献的大小;最后根据整体的编码差异来判别其维吾尔族人脸图像的全局特征和局部特性的表现来辨认最终结果。通过实验表明,本文算法有效的提高了在非均匀光照和局部遮挡下维吾尔族人脸图像时的识别效果,在非均匀光照和局部遮挡下的识别率分别达到了95%和90%以上,达到了较好的鲁棒性和实时性。
For non-uniform illumination and partial occlusion factors interfering Uygur recognitioneffect,affecting the problem Uyghur face recognition feature extraction efficiency and accuracy of theUighur proposed based on global features and local features combined with sparse coding Uyghur faceimage recognition algorithms. First, extract the global and local Uyghur face features an imagecomposition further multibank dictionary; then Uyghur face image feature vector construct correspondingjointly encoded in its inherent characteristics in the dictionary,then the different characteristics of thetwo-dimensional coding coefficient global features variance minimized mode,so that Uyghur face imageand the local feature exhibits inherent similarities,and by constructing different Uyghur face imagefeature coding coefficients and different from the right to characterize its image features a the size of thecontribution; Finally,according to the overall performance of its coding difference discriminates Uyghurface image global features and local characteristics to identify the final result. Experiments show that thealgorithm effectively improve the recognition results in non-uniform illumination and partial occlusion under Uyghur face image recognition rate under non-uniform illumination and partial occlusion,respectively,more than 95% and 90%,reaching robust and real-time.
出处
《电子设计工程》
2018年第2期50-55,60,共7页
Electronic Design Engineering
基金
国家自然科学基金(61462082)
