Scholarlink

<p>Although R offers a powerful environment for remote sensing analysis, it still lacks dedicated tools for the automated mapping of forest fire scars. To address this gap, we introduce OtsuSeg, an open-source R package that implements Otsu’s global thresholding algorithm for unsupervised burn scar segmentation and includes routines to assess classification accuracy using standard performance metrics. The package is designed to process pre- and post-fire optical imagery (Landsat, Sentinel-2) and streamline the extraction and evaluation of fire-affected areas. We applied OtsuSeg to four recent wildfire events in Tunisia and Algeria, validating the outputs against high-resolution reference perimeters from the Copernicus Emergency Management Service (EMS) Rapid Mapping. The results demonstrate that OtsuSeg consistently achieves high performance, with F1-scores exceeding 0.85 across all test cases. Moreover, a direct comparison with the binary thresholding tool in ArcGIS Pro—also based on Otsu’s method—revealed comparable accuracy, while highlighting OtsuSeg’s advantages in transparency, reproducibility, and accessibility. Beyond its practical utility for operational fire mapping, OtsuSeg also serves as an educational and research-oriented tool, offering a cost-effective solution for environmental monitoring and spatial data analysis.</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

OtsuSeg: an R package implementing Otsu’s thresholding technique for mapping forest fire scars with integrated accuracy assessment

  • 期刊论文

  • 发表日期：

    2025年8月20日

查看全文
  • Hammadi Achour,
  • Olga Viedma,
  • Zina Soltani,
  • Imene Habibi,
  • Wahbi Jaouadi

摘要

Although R offers a powerful environment for remote sensing analysis, it still lacks dedicated tools for the automated mapping of forest fire scars. To address this gap, we introduce OtsuSeg, an open-source R package that implements Otsu’s global thresholding algorithm for unsupervised burn scar segmentation and includes routines to assess classification accuracy using standard performance metrics. The package is designed to process pre- and post-fire optical imagery (Landsat, Sentinel-2) and streamline the extraction and evaluation of fire-affected areas. We applied OtsuSeg to four recent wildfire events in Tunisia and Algeria, validating the outputs against high-resolution reference perimeters from the Copernicus Emergency Management Service (EMS) Rapid Mapping. The results demonstrate that OtsuSeg consistently achieves high performance, with F1-scores exceeding 0.85 across all test cases. Moreover, a direct comparison with the binary thresholding tool in ArcGIS Pro—also based on Otsu’s method—revealed comparable accuracy, while highlighting OtsuSeg’s advantages in transparency, reproducibility, and accessibility. Beyond its practical utility for operational fire mapping, OtsuSeg also serves as an educational and research-oriented tool, offering a cost-effective solution for environmental monitoring and spatial data analysis.