Related Experiment Video
Updated: Sep 17, 2025
11:49
Cereal Crop Ear Counting in Field Conditions Using Zenithal RGB Images
Published on: February 2, 2019
9.4K
Evaluation of different spectral indices for wheat lodging assessment using machine learning algorithms.
Shikha Sharda1, Sumit Kumar1, Raj Setia2
1Punjab Remote Sensing Centre, Ludhiana, Punjab, India.
Scientific Reports
|July 2, 2025
Summary
Accurate wheat lodging detection using Sentinel-2 data and machine learning is vital for Indian agriculture. Random Forest with spectral indices like SSI and GDVI achieved 89.2% accuracy, aiding yield loss assessment.
Area of Science:
- Agricultural Remote Sensing
- Machine Learning Applications in Agriculture
- Crop Monitoring and Management
Background:
- Wheat lodging significantly impacts grain yield and harvest efficiency, necessitating precise assessment methods.
- Existing machine learning studies on wheat lodging lack comprehensive evaluation, especially for Indian agricultural conditions.
- Timely detection of wheat lodging is crucial for mitigating economic losses and improving crop management strategies.
Purpose of the Study:
- To systematically assess machine learning algorithms for detecting wheat lodging using multi-temporal Sentinel-2 data in Indian agricultural fields.
- To evaluate the performance of Random Forest (RF), Decision Tree (DT), and Support Vector Machine (SVM) for wheat classification and lodging detection.
- To identify optimal spectral indices and features for distinguishing lodged from non-lodged wheat.
Main Methods:
- Collected ground control points for healthy and lodged wheat areas in Ludhiana, Punjab.
- Analyzed temporal crop phenology and computed Normalized Difference Vegetation Index (NDVI) from November 2022 to April 2023.
- Implemented RF, DT, and SVM algorithms, and evaluated eight spectral indices (including SSI and GDVI) for lodging detection using Sentinel-2 data.
Main Results:
- Random Forest (RF) outperformed Decision Tree (DT) and Support Vector Machine (SVM) in wheat area extraction and classification accuracy.
- Spectral Sum Index (SSI) and Generalized Difference Vegetation Index (GDVI) were identified as key spectral indices for differentiating lodged from non-lodged wheat.
- The RF model, utilizing SSI and GDVI, achieved the highest overall accuracy of 89.2% for wheat lodging assessment.
Conclusions:
- The combination of Sentinel-2 derived SSI and GDVI with the Random Forest model provides an effective approach for spatio-temporal wheat lodging assessment.
- This methodology can support the development of decision support systems for accurate crop yield loss estimation in Indian agriculture.
- The study highlights the potential of advanced remote sensing and machine learning techniques for precision agriculture and crop insurance.
