基于局部约束编码的稀疏保持投影降维识别方法研究

稀疏表示技术的引入可有效解决降维处理对图参数的依赖，但这类降维方法不能同时兼顾稀疏重构和样本数据的邻近性问题。针对该问题，本文提出了一种基于局部约束编码的稀疏保持投影降维识别方法。通过稀疏表示分类模型构建了图边权矩阵，引入局部约束因子设计了降维投影模型，推导降维求解过程，分析了本文方法与SPP （ Sparse Preserving Projections ）和SLPP（ Soft Locality Preserving Projections ）方法之间的共性和区别，最后给出了识别算法流程。采用人脸图像数据集和高分辨SAR（ Synthetic Aperture Radar ）图像数据集对算法的有效性进行仿真验证，由于考虑了数据间的邻近性，本文方法较传统方法可获得更好的识别性能。

Sparsity Preserving Projections Based on Locality Constrained Coding with Applications for Targets Recognition

Constructing graph by sparse representation ( SP) can reduce the dimensionality reduction ( DR) which re-lies on neighborhood parameter selection.However,these DR algorithms are usually unable to take sparse reconstruction into consideration while preserving local data structure.This paper presents a sparsity preserving projections based on locality-constrained coding ( LCC-SPP) algorithm.Firstly,an“adjacent” weight matrix of dataset is constructed by sparse represen-tation based classification ( SRC) .Then,a locality adaptor is introduced and the dimension reduction is modeled.We derive the solution of objective function.The similarities and differences are presented with sparse preserving projections ( SPP ) and soft locality preserving projections ( SLPP) ,respectively.At last,the recognition flow is given.We conduct experiments on databases designed for face and synthetic aperture radar ( SAR) images recognition.Considering the data locality,the pro-posed method has better recognition performance than SPP and SLPP.