Scholarlink

Image classification under challenging conditions such as extreme weather, variable lighting, and thermal distortions has become increasingly critical, especially where conventional visual imaging fails. This study proposes an ensemble approach combining Random Forest, XGBoost, and Support Vector Machine (SVM) classifiers to enhance robustness in noisy environments. The model is designed to address interference from fog, fluctuating temperatures, and inconsistent lighting. Compared to a baseline Convolutional Neural Network (CNN), the ensemble achieves superior performance, with 96.14% accuracy, 96.28% precision, 96.07% recall, and a 96.00% F1-score. A confusion matrix further confirms its reliability. This approach is aimed at improving the reliability of thermal imaging systems used in critical applications like industrial monitoring, search and rescue, and surveillance by addressing noise interference and environmental fluctuation.

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

Thermal Image Classification Under Adverse Conditions Using Ensemble Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Akash Choudhary,
  • Richa Dhiman,
  • Deepika Rani,
  • Aditya Kumar Singh

摘要

Image classification under challenging conditions such as extreme weather, variable lighting, and thermal distortions has become increasingly critical, especially where conventional visual imaging fails. This study proposes an ensemble approach combining Random Forest, XGBoost, and Support Vector Machine (SVM) classifiers to enhance robustness in noisy environments. The model is designed to address interference from fog, fluctuating temperatures, and inconsistent lighting. Compared to a baseline Convolutional Neural Network (CNN), the ensemble achieves superior performance, with 96.14% accuracy, 96.28% precision, 96.07% recall, and a 96.00% F1-score. A confusion matrix further confirms its reliability. This approach is aimed at improving the reliability of thermal imaging systems used in critical applications like industrial monitoring, search and rescue, and surveillance by addressing noise interference and environmental fluctuation.