Comparison of Hyperspectral Atmospheric Correction Algorithms for Precise Mapping of Rice Crop

Abstract

For millions of people, rice means life, and therefore, it is harvested in many regions of the world. Two rice species are primarily cultivated in the world, namely Asian and African rice. It grows primarily in major river deltas, such as Asia and Southeast Asia. Conventional method of mapping rice crop area is tedious and time-consuming job and more often subjected to erroneous results. In this study advanced remote sensing technique is used for mapping, to map precisely hyperspectral remote sensing with different atmospheric algorithms were compared for better accuracy. Also different supervised classification techniques were compared for the accurate area mapping of rice crop. The ASD field spec Pro hand held spectroradiometer is used for reference spectra collection. And high accuracy GPS device is used to collect ground truth information. Results show that both FLAASH and HAC algorithms produce a good spectrum with respect to the rice spectra obtained from ASD handheld spectroradiometer. SAM image classification and Parallelepiped classifier were used for classification of imagery. From the accuracy assessment performed, accuracy of 88% by using SAM and 84% obtained using Parallelepiped classifier for Hooghly region and 93% using SAM and 87% using Parallelepiped for West Godavari region. From the study, it was found that the best approach for rice crop mapping in Hooghly and West Godavari is SAM classification. The study helps to map the rice crop area accurately; it can be used for yield estimation, indirectly which is helpful for policy makers and to estimate the export, import potential. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.