Scholarlink

<p>Various objects on Earth can be classified using the hundreds of contiguous spectral bands that comprise Hyperspectral Image (HSI). However, the Hughes phenomenon and high processing computational cost result from the continuous narrow bands with high spectral correlation. By eliminating redundant bands, band selection has been shown to be efficient in preventing such issues. Further, performance of HSI classification can be improved by employing a metaheuristic technique with high exploration and exploitation capabilities for band selection. Here, a new HSI classification model is developed based on hybrid deep learning named ResNeXt Maxout Network (ResNeXtMN), which is created by incorporating the ResNeXt and Deep Maxout Network (DMN). Here, the selection of the optimal band for the HSI classification is done using proposed Adam Drawer Algorithm (ADA) that is formulated using Adam Optimizer and Drawer Algorithm (DA). This optimization technique provides the optimal band selected as a solution variable for the effective HSI classification based on the proposed ResNeXtMN. The effectiveness of the ResNeXtMN model is assessed by specificity, accuracy and sensitivity and it gained extreme values such as 92.8%, 90.9%, and 94.9%, respectively.</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

Enhanced hyperspectral image classification with optimized band selection and Resnext maxout network

  • 期刊论文

  • 发表日期：

    2025年8月6日

查看全文
  • Sheela Jayachandran,
  • Aparna Shivampeta,
  • Chaya Ravindra,
  • Mohammed Abdulmajeed Moharram,
  • Divya Meena Sundaram,
  • Sachi Nandan Mohanty

摘要

Various objects on Earth can be classified using the hundreds of contiguous spectral bands that comprise Hyperspectral Image (HSI). However, the Hughes phenomenon and high processing computational cost result from the continuous narrow bands with high spectral correlation. By eliminating redundant bands, band selection has been shown to be efficient in preventing such issues. Further, performance of HSI classification can be improved by employing a metaheuristic technique with high exploration and exploitation capabilities for band selection. Here, a new HSI classification model is developed based on hybrid deep learning named ResNeXt Maxout Network (ResNeXtMN), which is created by incorporating the ResNeXt and Deep Maxout Network (DMN). Here, the selection of the optimal band for the HSI classification is done using proposed Adam Drawer Algorithm (ADA) that is formulated using Adam Optimizer and Drawer Algorithm (DA). This optimization technique provides the optimal band selected as a solution variable for the effective HSI classification based on the proposed ResNeXtMN. The effectiveness of the ResNeXtMN model is assessed by specificity, accuracy and sensitivity and it gained extreme values such as 92.8%, 90.9%, and 94.9%, respectively.