Open-source hardware platform for automated, non-invasive plant phenotyping through combined RGB and thermal imaging.

We present an open-source, modular plant phenotyping platform for non-invasive monitoring of small to medium-sized plants in growth chambers. The system integrates an RGB camera and a FLIR Lepton 3.5 thermal module on an automated two-axis Cartesian gantry capturing overhead images at 24 predefined positions. A key design feature is the use of Hall effect sensors with permanent magnets for discrete position referencing, providing robust, recalibration-free positioning without complex mechanical systems such as lead screws, belt-driven encoders, or step-counting alone. The platform uses 3D-printed components, a Raspberry Pi, stepper motors with rack-and-pinion transmission, and commonly available hardware store materials. Automated image acquisition scheduling increases temporal resolution and eliminates errors from prolonged manual measurements. Companion software enables homography-based projection of RGB segmentation masks onto thermal images and interactive exploration and quantitative analysis of thermographic data. As proof of concept, a multi-day water deficit experiment with soybean plants was conducted. RGB images were segmented using a U-Net deep learning model, and projected leaf area dynamics clearly discriminated between irrigated and drought-stressed plants. Thermal imaging revealed canopy temperature differences consistent with stomatal closure under water deficit.