融合约束采样和面部对齐的稀疏表示人脸识别

针对传统的稀疏表示分类算法中面部对齐受限而影响人脸识别率的问题,提出一种基于约束采样和面部对齐的稀疏表示分类算法。首先通过使用约束采样对训练图像进行预先标注得到固定脸特征;然后结合图像的纹理信息和形状特征进行面部对齐及特征提取;最后计算出测试样本与各个训练样本之间的相似度,利用稀疏表示分类器完成人脸的识别。在AR、CAS-PEAL及扩展YaleB人脸数据库上的实验验证了算法的有效性及鲁棒性。实验结果表明,约束采样和面部对齐的组合大大提高了人脸识别率,相比几种较为先进的鲁棒人脸识别算法,该算法取得了更好的识别效果。

通用学习框架结合2DPCA的鲁棒人脸识别

针对现有的人脸识别算法由于光照、表情、姿态、伪装等变化而严重影响识别性能的问题,提出了一种基于通用学习框架结合2DPCA的鲁棒人脸识别算法。首先借助于额外的通用训练样本集进行样本的叠加以增加训练样本的数量;然后利用经典的2DPCA算法进行特征提取;最后,利用最近邻分类器对人脸进行分类并完成最终的人脸识别。在基准人脸数据库ORL、FERET及鲁棒人脸数据库AR、扩展YaleB上的实验验证了该算法的有效性及鲁棒性,实验结果表明,相比其他几种人脸识别算法,提出的算法不仅提高了人脸识别率,而且大大地减少了识别所用时间,有望应用于实时鲁棒人脸自动识别系统中。

Does disruptive camouflage conceal edges and features？

Camouflage is ubiquitous in the natural world and benefits both predators and prey. Amongst the range of conceal- ment strategies, disruptive coloration is thought to visually fragment an animal＇s＇ outline, thereby reducing its rate of discovery. Here, I propose two non-mutually exclusive hypotheses for how disruptive camouflage functions, and describe the visual me- chanisms that might underlie them. （1） The local edge disruption hypothesis states that camouflage is achieved by breaking up edge information. （2） The global feature disruption hypothesis states camouflage is achieved by breaking up the characteristic features of an animal （e.g., overall shape or facial features）. Research clearly shows that putatively disruptive edge markings do increase concealment; however, few tests have been undertaken to determine whether this survival advantage is attributable to the distortion of features, so the global feature disruption hypothesis is under studied. In this review the evidence for global feature disruption is evaluated. Further, I address if object recognition processing provides a feasible mechanism for animals＇ features to influence concealment. This review concludes that additional studies are needed to test if disruptive camouflage operates through the global feature disruption and proposes future research directions [Current Zoology 61 （4）： 708-717, 2015].