Scholarlink

As one of the staple horticultural crops, Shallot requires accurate monitoring to understand its cultivation better and to face environmental challenges in lowland regions. Nevertheless, there are still limited studies on mapping shallot agriculture using remote sensing data. In this study, we aimed to enhance the accuracy and efficiency of mapping shallot cultivation in lowland areas by leveraging high-resolution Sentinel-2 imagery combined with machine-learning techniques, specifically Random Forest, to classify the cultivation areas accurately. Preprocessed Sentinel-2 data using Google Earth Engine and validation data from field surveys and high-resolution satellite imagery inspection are utilized to create training and testing datasets for the machine learning model. Our findings reveal that the Random Forest model resulted in an Overall Accuracy of 88% and a total area of 12,121 hectares in July 2023. The proposed method can be used to estimate the planting and harvested area regularly to support smart-climate agriculture.

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

Mapping Shallot in Lowland Areas Through Sentinel-2 Imagery and Machine Learning Approach

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Robby Arifandri,
  • Vicca Karolinoerita,
  • Fadhlullah Ramadhani,
  • Dede Dirgahayu,
  • Lukita Devy,
  • Muhammad Iqbal Habibie,
  • Yudi Riadi Fanggidae,
  • Dharmayanti,
  • Robby Kurniawan Harahap,
  • Swelandiah Endah Pratiwi,
  • Widiastuti

摘要

As one of the staple horticultural crops, Shallot requires accurate monitoring to understand its cultivation better and to face environmental challenges in lowland regions. Nevertheless, there are still limited studies on mapping shallot agriculture using remote sensing data. In this study, we aimed to enhance the accuracy and efficiency of mapping shallot cultivation in lowland areas by leveraging high-resolution Sentinel-2 imagery combined with machine-learning techniques, specifically Random Forest, to classify the cultivation areas accurately. Preprocessed Sentinel-2 data using Google Earth Engine and validation data from field surveys and high-resolution satellite imagery inspection are utilized to create training and testing datasets for the machine learning model. Our findings reveal that the Random Forest model resulted in an Overall Accuracy of 88% and a total area of 12,121 hectares in July 2023. The proposed method can be used to estimate the planting and harvested area regularly to support smart-climate agriculture.