Scholarlink

<p>In the domain of image-set analysis, the Grassmann Manifold serves as an efficacious instrument, though the computational expenses it incurs are notably substantial. Locality Preserving Projection on Grassmann Manifold (GLPP), a classical algorithm for dimensionality reduction, effectively alleviates these computational burdens. However, it is important to note that the construction of the neighborhood graph in GLPP overlooks the categorical information of the data, thereby limiting its discriminative capacity. To address this issue, we propose the Discriminant Locality Preserving Projection on Grassmann Manifold (GDLPP) and Semi-Supervised Locality Preserving Projection on Grassmann Manifold (GSLPP). Classification outcomes across 6 image-set datasets reveal that the classification capabilities of GDLPP and GSLPP surpass those of existing image-set classification algorithms. By effectively preserving the local manifold structure of the data and comprehensively leveraging the label information, GDLPP and GSLPP notably augment the feature extraction and classification prowess of GLPP.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Discriminant locality preserving projection on Grassmann Manifold for image-set classification

  • 期刊论文

  • 发表日期：

    2025年1月17日

查看全文
  • Benchao Li,
  • Ting Wang,
  • Ruisheng Ran

摘要

In the domain of image-set analysis, the Grassmann Manifold serves as an efficacious instrument, though the computational expenses it incurs are notably substantial. Locality Preserving Projection on Grassmann Manifold (GLPP), a classical algorithm for dimensionality reduction, effectively alleviates these computational burdens. However, it is important to note that the construction of the neighborhood graph in GLPP overlooks the categorical information of the data, thereby limiting its discriminative capacity. To address this issue, we propose the Discriminant Locality Preserving Projection on Grassmann Manifold (GDLPP) and Semi-Supervised Locality Preserving Projection on Grassmann Manifold (GSLPP). Classification outcomes across 6 image-set datasets reveal that the classification capabilities of GDLPP and GSLPP surpass those of existing image-set classification algorithms. By effectively preserving the local manifold structure of the data and comprehensively leveraging the label information, GDLPP and GSLPP notably augment the feature extraction and classification prowess of GLPP.