Deep Learning for Soil Moisture Content Estimation via Reflectance Multispectral Imaging
Lists
Only you see this
Abstract
Soil moisture content estimation is essential for soil quality assurance before undertaking construction operations. Existing methodologies for estimating soil moisture content are often associated with significant costs and considerable time requirements. This study introduces a pragmatic, cost-effective, accurate, and prompt indirect method for estimating soil moisture content using a multispectral imaging based regression neural network. An in-house built, site-testing capable multispectral imaging system operating in the range from 365 nm to 940 nm is used to obtain multispectral images of soil samples. Image preprocessing techniques such as dark current subtraction and histogram equalization are used to mitigate the random noise and surface imperfections in the images. A regression neural network is utilized to estimate the soil moisture content with R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value of 0.9987. Further testing with validation data gives R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value of 0.9922 ensuring the suitability of this method for real-world applications.
BibTeX
@misc{ranasinghe2024deep,
author = {Subhani Ranasinghe and Shiran Jayakody and Mario De Silva and Vijitha Herath and Roshan Godaliyadda and M. P. B. Ekanayake and S. K. Navaratnarajah and Fadi Kizel and Gayathri Thilakarathne},
title = {{Deep Learning for Soil Moisture Content Estimation via Reflectance Multispectral Imaging}},
year = {2024},
doi = {10.1109/icatc64549.2024.11025290},
}Loading related content...