Scholarlink

Accurate urban vegetation inventorying is essential for effective landscape management, and remote sensing data, like VHR imagery, aids in vegetation extraction. This paper introduces a method for land use classification using multi-temporal remote sensing data and semi-supervised learning. A random forest ensemble serves as the base classifier, utilizing probability distributions to address spectral variations across time phases. A cascade mechanism enables classifier communication across phases, while joint confidence maps improve unlabeled sample selection and classification accuracy. Tested on SENTINEL-2-MSI images of Nanjing City, this approach shows enhanced accuracy and consistency in multi-temporal land use classification, especially for vegetation.

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

Multi-temporal Remote Sensing Image Classification Using Semi-supervised Learning and Random Forest Ensemble

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rishav Shrivastava

摘要

Accurate urban vegetation inventorying is essential for effective landscape management, and remote sensing data, like VHR imagery, aids in vegetation extraction. This paper introduces a method for land use classification using multi-temporal remote sensing data and semi-supervised learning. A random forest ensemble serves as the base classifier, utilizing probability distributions to address spectral variations across time phases. A cascade mechanism enables classifier communication across phases, while joint confidence maps improve unlabeled sample selection and classification accuracy. Tested on SENTINEL-2-MSI images of Nanjing City, this approach shows enhanced accuracy and consistency in multi-temporal land use classification, especially for vegetation.