Scholarlink

<p>The escalating demand for green fodder, driven by the adoption of high-yielding livestock breeds and expansion of cooperative dairy farming, necessitates robust methods for monitoring fodder crop dynamics. Synthetic Aperture Radar (SAR), with its all-weather, day and night imaging capability, presents a promising tool for crop discrimination and forecasting. This study investigates the potential of time-series dual-polarized Sentinel-1&#xa0;A SAR data for discriminating <i>Kharif</i> season fodder and non-fodder crops in three blocks of Banaskantha district, Gujarat, India, during July–October 2021. Polarimetric decomposition using the H/A/Alpha method was applied to derive temporal profiles of entropy, anisotropy, and alpha angle. Additionally, the Polarimetric Radar Vegetation Index (PRVI) was computed from Stoke’s parameters. Random Forest (RF) classifier which is an ensemble machine learning algorithm was employed to assess crop separability based on VV and VH backscatter intensities and derived polarimetric features. The temporal backscatter patterns varied distinctly across crop types: multi-cut fodder crops (e.g., napier, bajra) exhibited periodic peaks post-harvest, short-duration crops (e.g., cowpea) showed rapid rise and decline, while long-duration non-fodder crops (e.g., castor, groundnut) displayed gradual backscatter increase, peaking during vegetative to reproductive phases. Classification using combined polarimetric parameters (alpha, entropy, anisotropy) achieved the highest accuracy (93.5%) with a Kappa coefficient of 0.90, outperforming PRVI (90.2%, Kappa = 0.87) and amplitude-based features (81.6%). The results underscore the efficacy of integrating temporal polarimetric SAR features with machine learning for accurate, timely discrimination of fodder crops, thereby enhancing decision-making in fodder resource management.</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

Mapping and Monitoring of Kharif Fodder crops using Temporal Polarimetric SAR Features and Random Forest

  • 期刊论文

  • 发表日期：

    2026年3月31日

查看全文
  • Mukesh Kumar,
  • Saroj Maity,
  • Sujay Dutta,
  • Bimal K. Bhattacharya

摘要

The escalating demand for green fodder, driven by the adoption of high-yielding livestock breeds and expansion of cooperative dairy farming, necessitates robust methods for monitoring fodder crop dynamics. Synthetic Aperture Radar (SAR), with its all-weather, day and night imaging capability, presents a promising tool for crop discrimination and forecasting. This study investigates the potential of time-series dual-polarized Sentinel-1 A SAR data for discriminating Kharif season fodder and non-fodder crops in three blocks of Banaskantha district, Gujarat, India, during July–October 2021. Polarimetric decomposition using the H/A/Alpha method was applied to derive temporal profiles of entropy, anisotropy, and alpha angle. Additionally, the Polarimetric Radar Vegetation Index (PRVI) was computed from Stoke’s parameters. Random Forest (RF) classifier which is an ensemble machine learning algorithm was employed to assess crop separability based on VV and VH backscatter intensities and derived polarimetric features. The temporal backscatter patterns varied distinctly across crop types: multi-cut fodder crops (e.g., napier, bajra) exhibited periodic peaks post-harvest, short-duration crops (e.g., cowpea) showed rapid rise and decline, while long-duration non-fodder crops (e.g., castor, groundnut) displayed gradual backscatter increase, peaking during vegetative to reproductive phases. Classification using combined polarimetric parameters (alpha, entropy, anisotropy) achieved the highest accuracy (93.5%) with a Kappa coefficient of 0.90, outperforming PRVI (90.2%, Kappa = 0.87) and amplitude-based features (81.6%). The results underscore the efficacy of integrating temporal polarimetric SAR features with machine learning for accurate, timely discrimination of fodder crops, thereby enhancing decision-making in fodder resource management.