Scholarlink

The satellite pictures with hyperspectral imagery (HSI) will be worked on more so that they can be used in any way. In order to get the map's coordinates from picture coordinates, the most important and necessary step is to classify hyperspectral images. In this method, the ground control points (GCPs) must be physically removed from the images collected distantly using the ground truth values. This takes time. In this study, super-pixel-based principal component analysis (SuperPCA) is offered as a way to classify the multitemporal HSI satellite data. This approach would be automatically find GCPs and cut down on the time it takes to process, which will both improve accuracy. PCA is a famous algorithm for extracting features from satellite data, although it takes a long time to do so. To get around the problems with PCA, SuperPCA has been proposed and used only for pictures with fewer features. It has not been used on satellite images. In this study, the phase angle of the SuperPCA is changed so that it can be used with satellite data to find features at six levels in multitemporal HSI satellite imagery. Support vector machine (SVM) is often worn for multi-class segmentation that is not linear. The kernel decision is what makes SVM work so well. So, fuzzy-SVM (F-RVM) is suggested for kernel choosing depending on the degree of detail and extent of the features in the satellite imagery. The outcomes are evaluating to standard methods and show that the new method works better.

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

SuperPCA-Based Machine Learning System by Hyperspectral Image Classification Assessment Using SVM

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • V. A. Narayana,
  • R. Venkateswara Reddy,
  • V. Venkataiah,
  • K. Srujan Raju,
  • Mohd. Abdul Naqi,
  • Syeda Sumaiya Afreen

摘要

The satellite pictures with hyperspectral imagery (HSI) will be worked on more so that they can be used in any way. In order to get the map's coordinates from picture coordinates, the most important and necessary step is to classify hyperspectral images. In this method, the ground control points (GCPs) must be physically removed from the images collected distantly using the ground truth values. This takes time. In this study, super-pixel-based principal component analysis (SuperPCA) is offered as a way to classify the multitemporal HSI satellite data. This approach would be automatically find GCPs and cut down on the time it takes to process, which will both improve accuracy. PCA is a famous algorithm for extracting features from satellite data, although it takes a long time to do so. To get around the problems with PCA, SuperPCA has been proposed and used only for pictures with fewer features. It has not been used on satellite images. In this study, the phase angle of the SuperPCA is changed so that it can be used with satellite data to find features at six levels in multitemporal HSI satellite imagery. Support vector machine (SVM) is often worn for multi-class segmentation that is not linear. The kernel decision is what makes SVM work so well. So, fuzzy-SVM (F-RVM) is suggested for kernel choosing depending on the degree of detail and extent of the features in the satellite imagery. The outcomes are evaluating to standard methods and show that the new method works better.