Scholarlink

<p>Government of India strategic studies rightly mentions that there exist technological gaps in mineral exploration, particularly with respect to hyperspectral remote sensing to detect and characterize mineralized hotspots. In the present study, classification of EO-1 Hyperion (hyperspectral) data has been attempted using the SAM algorithm of ENVI and several ML techniques as a novel integrated approach dedicated to the detection and mapping of iron ore minerals from open cast projects in Northern Odisha. Spectrally pure pixels available in the Top of the Atmosphere corrected reflectance Hyperion image were devolved into three major endmembers, viz. iron ore mines and fallow lands. The employed ML techniques indicated that XGBoost, Random Forest, and SMOTE-RF provided the most accurate predictions for the&#xa0;detection of iron ore mines, whereas LightGBM and Naïve Bayes showed some uncertainties. This technique, in a holistic way, holds promise for the automatic detection of iron ore mines in relatively unexplored but potentially enriched mineralized zones. </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

Synergetic Usage of Hyperspectral Remote Sensing and Machine Learning in Detection and Mapping of Iron Ore Deposit Hotspots: A Case Study from Northern Odisha, India

  • 期刊论文

  • 发表日期：

    2025年7月12日

查看全文
  • S. Aditya Narayanan,
  • Sudikin Pramanik

摘要

Government of India strategic studies rightly mentions that there exist technological gaps in mineral exploration, particularly with respect to hyperspectral remote sensing to detect and characterize mineralized hotspots. In the present study, classification of EO-1 Hyperion (hyperspectral) data has been attempted using the SAM algorithm of ENVI and several ML techniques as a novel integrated approach dedicated to the detection and mapping of iron ore minerals from open cast projects in Northern Odisha. Spectrally pure pixels available in the Top of the Atmosphere corrected reflectance Hyperion image were devolved into three major endmembers, viz. iron ore mines and fallow lands. The employed ML techniques indicated that XGBoost, Random Forest, and SMOTE-RF provided the most accurate predictions for the detection of iron ore mines, whereas LightGBM and Naïve Bayes showed some uncertainties. This technique, in a holistic way, holds promise for the automatic detection of iron ore mines in relatively unexplored but potentially enriched mineralized zones.