MoistureMapper: Non-Invasive Soil Moisture Detection with a Millimeter-Wave Sensor

TLDR

• Built a radar-based, non-invasive moisture embedded sensing system using TI AWR1642 mmWave + GPS.
• Processed stabilized radar returns in Python and classified soil moisture into four discrete levels with 95%+ accuracy.
• Results and system summary are shown in the poster below.

Overview

MoistureMapper estimates soil moisture using 77 GHz reflections correlated with soil permittivity. The goal is a low-cost, non-invasive alternative to probe-based measurements for golf-course greens and similar settings.

Hardware & System

• Sensor: TI AWR1642 (1 Tx / 1 Rx prototype configuration).
• Controller/Logging: SparkFun RedBoard, geotagging via Adafruit PA1010D GPS.
• Pipeline: Python feature extraction from radar measurements → classification of moisture level.

Method

• Collected radar frames over sand/soil at controlled moisture levels.
• Engineered features (e.g., band energies, peak magnitude, decay slope, phase statistics).
• Trained a lightweight model to classify four moisture levels from radar features.

Contributions & Acknowledgements

Built in collaboration with Eli Propp, Tanzim Pathan, and Derick Jones.