基于特征优化与稀疏表示的3D掌纹分类

3D palm-print sparse representation classification based on optimized features

针对大数据背景下3D掌纹技术存在的问题,提出一种基于优化投影矩阵的3D掌纹稀疏表示识别技术架构。系统首先提取3D掌纹表面类型特征,然后利用分块方向梯度直方图构成训练样本,通过优化设计投影矩阵,使得同类掌纹投影特征互相关性变大,异类掌纹投影特征互相关性变小;最后利用投影后3D掌纹特征稀疏表示分类,并比较L0/L1/L2范数各种快速算法性能。通过投影优化后的系统,在识别率和实时性上都有所改善,仿真实验证实了研究工作的有效性。

In response to the 3 Dpalm-print technical problems in the context of big data,this paper proposes a new 3 D palm-print sparse representation classification technique based on optimized features.Firstly,the 3 Dpalm-print surface pattern features are extracted from the training images.Secondly,the blocked histogram oriented gradient features are vectorized as the training samples.A projection matrix is optimized from the training features,which enhance the homogeneous coherence and reduces the heterogeneous one.Finally,the optimized features are used in the sparse representation classification system via L0/L1/L2 norm representation algorithm through comparison.With these modifications,the new 3 Dpalm-print recognition system can improve real-time performance and recognition rate greatly.Experiments results validate the proposed methods.