A novel coding based method named as local binary orientation code （LBOCode） for palmprint recognition is proposed. The palmprint image is firstly convolved with a bank of Gabor filters, and then the orientation inf...A novel coding based method named as local binary orientation code （LBOCode） for palmprint recognition is proposed. The palmprint image is firstly convolved with a bank of Gabor filters, and then the orientation information is attained with a winner-take-all rule. Subsequently, the resulting orientation mapping array is operated by uniform local binary pattern. Accordingly, LBOCode image is achieved which contains palmprint orientation information in pixel level. Further we divide the LBOCode image into several equal-size and nonoverlapping regions, and extract the statistical code histogram from each region independently, which builds a global description of palmprint in regional level. In matching stage, the matching score between two palmprints is achieved by calculating the two spatial enhanced histograms＇ dissimilarity, which brings the benefit of computational simplicity. Experimental results demonstrate that the proposed method achieves more promising recognition performance compared with that of several state-of-the-art methods.展开更多
mean shift目标跟踪使用任一个单独特征都很难对大幅度的光照、背景变化和车辆大范围运动具有强鲁棒性,如单一的颜色特征对光照较为敏感,光照突变容易导致目标丢失。当背景颜色和目标颜色相近时也易造成目标丢失,因此利用目标的单一特...mean shift目标跟踪使用任一个单独特征都很难对大幅度的光照、背景变化和车辆大范围运动具有强鲁棒性,如单一的颜色特征对光照较为敏感,光照突变容易导致目标丢失。当背景颜色和目标颜色相近时也易造成目标丢失,因此利用目标的单一特征信息往往很难实现鲁棒的目标跟踪。文章提出基于颜色和LBP多特征mean shift跟踪方法,使跟踪结果不再过分依赖某一特征,增强了对背景变化、目标大范围运动的鲁棒性。展开更多
微表情是一个人试图隐藏内心真实情感却又不由自主流露出的不易被察觉的面部表情。与一般面部表情相比,微表情最显著的特点是持续时间短、强度弱,往往难以有效识别。文中提出了一种基于LBP-TOP(Local Binary Pattern from Three Orthogo...微表情是一个人试图隐藏内心真实情感却又不由自主流露出的不易被察觉的面部表情。与一般面部表情相比,微表情最显著的特点是持续时间短、强度弱,往往难以有效识别。文中提出了一种基于LBP-TOP(Local Binary Pattern from Three Orthogonal Planes)特征和支持向量机(Support Vector Machine,SVM)分类器的微表情识别方法。首先,采用LBP-TOP算子来提取微表情特征;然后,提出一种基于ReliefF与局部线性嵌入(Locally Linear Embedding,LLE)流形学习算法相结合的特征选择算法,对提取的LBP-TOP特征向量进行降维;最后,使用径向基函数(Radial Basis Function,RBF)核的SVM分类器进行分类,将测试样本图像序列的微表情分为5类:高兴、厌恶、压抑、惊讶、其他。在CASME Ⅱ微表情数据库上采用"留一人交叉验证"(Leave-One-Subject-Out Cross Validation,LOSO-CV)的方式进行了实验,可得到58.98%的分类准确率。实验结果表明了该算法的有效性。展开更多
行人检测是目标识别领域的一大难题,针对行人检测存在特征维度高、检测耗时和精度低等问题,文章提出使用多尺度分块方式将样本图片在3个尺度下分别分割成5个区域,在每个区域中根据行人轮廓置信模板和梯度方向量化权值进行二次加权统计...行人检测是目标识别领域的一大难题,针对行人检测存在特征维度高、检测耗时和精度低等问题,文章提出使用多尺度分块方式将样本图片在3个尺度下分别分割成5个区域,在每个区域中根据行人轮廓置信模板和梯度方向量化权值进行二次加权统计得到梯度直方图(histogram of oriented gradient,HOG),并将其与Sobel边缘局部二元模式(Sobel edge local binary pattern,Sobel-LBP)算法相融合作为特征,然后采用线性支持向量机(support vector machine,SVM)分类方法学习得到行人检测分类器,最后使用滑动窗口法检测出行人。在MIT和INRIA库上的实验证明,该特征在学习和检测速度上都比HOG等方法有明显优势,能有效、准确、快速地检测行人。展开更多
基金supported partly by the National Grand Fundamental Research 973 Program of China under Grant No. 2004CB318005the Doctoral Candidate Outstanding Innovation Foundation under Grant No.141092522the Fundamental Research Funds under Grant No.2009YJS025
文摘A novel coding based method named as local binary orientation code （LBOCode） for palmprint recognition is proposed. The palmprint image is firstly convolved with a bank of Gabor filters, and then the orientation information is attained with a winner-take-all rule. Subsequently, the resulting orientation mapping array is operated by uniform local binary pattern. Accordingly, LBOCode image is achieved which contains palmprint orientation information in pixel level. Further we divide the LBOCode image into several equal-size and nonoverlapping regions, and extract the statistical code histogram from each region independently, which builds a global description of palmprint in regional level. In matching stage, the matching score between two palmprints is achieved by calculating the two spatial enhanced histograms＇ dissimilarity, which brings the benefit of computational simplicity. Experimental results demonstrate that the proposed method achieves more promising recognition performance compared with that of several state-of-the-art methods.
文摘行人检测是目标识别领域的一大难题,针对行人检测存在特征维度高、检测耗时和精度低等问题,文章提出使用多尺度分块方式将样本图片在3个尺度下分别分割成5个区域,在每个区域中根据行人轮廓置信模板和梯度方向量化权值进行二次加权统计得到梯度直方图(histogram of oriented gradient,HOG),并将其与Sobel边缘局部二元模式(Sobel edge local binary pattern,Sobel-LBP)算法相融合作为特征,然后采用线性支持向量机(support vector machine,SVM)分类方法学习得到行人检测分类器,最后使用滑动窗口法检测出行人。在MIT和INRIA库上的实验证明,该特征在学习和检测速度上都比HOG等方法有明显优势,能有效、准确、快速地检测行人。
