Development of microwave remote sensing applications for soil moisture estimation

• Main Author: Uthayakumar Akileshwaran
• Other Authors: Muhammad Faeyz Karim
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering::Antennas, wave guides, microwaves, radar, radio
• Online Access: https://hdl.handle.net/10356/142889
• Institution: Nanyang Technological University

Soil moisture is an important parameter under study to monitor both the chemical and physical properties of the soil. Currently, the soil moisture is measured either by contact sensing or remote sensing methods. The contact type sensors have various limitations in terms of continuous measurement and hence resulting in inaccurate measurements. Remote sensing method predominantly involves the use of microwave equipment for the measurement of soil moisture. Remote sensing utilizes the help of satellites which has created extensive research opportunities in this field. However, there has always been some disagreements between the actual and measured values from active microwave remote sensing methods. To overcome this issue, we intend to apply the concept of machine learning with the microwave based remote sensing data. The key objectives of this research work are to apply machine learning techniques to VNA based transmission method and use Walabot, a short-range RADAR to estimate soil moisture. In the first method, the transmission technique was used to determine the dielectric constant and Topps Equation was used to calculate the volumetric moisture content. Using this data, various machine learning models are studied, trained and tested to select the most-accurate model for this application. In the second approach, a Frequency Modulated Continuous Wave Radar operating in the frequency range 3.3 – 10.3 GHz, Walabot, has been used for soil moisture measurement. It is carried out with an intention to develop a product that is cost-effective for common users such as agriculturists to monitor the soil conditions with improved accuracy. In the machine learning approach for the transmission method, the Medium Gaussian SVM model was able to predict the soil moisture accurately with a R-squared percentage of 99.84%. The application developed for Walabot radar had given good results for dry soil and close to accurate results for moist soil.