Record Details

Machine Learning Approaches and Sentinel-2 Data in Crop Type Mapping

OAR@ICRISAT

View Archive Info
 
 
Field Value
 
Relation http://oar.icrisat.org/12138/
https://link.springer.com/chapter/10.1007/978-981-16-5847-1_8
https://doi.org/10.1007/978-981-16-5847-1_8
 
Title Machine Learning Approaches and Sentinel-2 Data in Crop Type
Mapping
 
Creator Panjala, P
Gumma, M K
Teluguntla, P
 
Subject Agriculture
Drylands Agriculture
 
Description Crop monitoring becomes essential in attaining food security for implementation of various agricultural serving programs. So, fast and reliable crop monitoring is must. Using traditional methods, crop monitoring maps need high amount
of satellite data downloading and processing time. Google Earth Engine (GEE) cloud platform enables us to save time in downloading and processing of time series satellite data, the every satellite imagery is converted into Normalized Difference Vegetation Index (NDVI) image and stacked monthlywisemaximum images. The stacked image was used for conducting supervised classification. The main objective of this study is to evaluate the performance of different supervised machine learning (ML) classifiers in GEE platform and Spectral Matching Technique (SMT) using Sentinel-2
10 m satellite imagery in specific crop type classification. The crop classification for the year 2018–19 (rabi season) was carried for Jhansi District using supervised classifiers like Random Forest (RF), Support Vector Machine (SVM) and Classification and Regression Trees (CART) in GEE platform and also with SMT with the help of ground data. It was attained nearly 81.8% accuracy for RF, 68.8% for SVM, 64.9% for CART and 88% for SMT. The results obtained using RF classifier were nearly relative to SMT classification map. The study indicates that classifier’s performance depends on the quality of ground data used, RF can reduce the error
samples in ground samples and produce satisfactory results. This study compared results obtained from all the above classifiers with agricultural statistics and also compared crop-wise accuracies. In the study, it was observed that RF classification is outperformed when compared with other classifiers considered in the study.
 
Publisher Springer
 
Contributor Reddy, G P O
Raval, M S
Adinarayana, G
Chaudhary, S
 
Date 2021-10-11
 
Type Book Section
PeerReviewed
 
Identifier Panjala, P and Gumma, M K and Teluguntla, P (2021) Machine Learning Approaches and Sentinel-2 Data in Crop Type Mapping. In: Data Science in Agriculture and Natural Resource Management. Studies in Big Data, 96 . Springer, Singapore, pp. 161-180. ISBN 2197-6503