Technical approach

The platform integrates complementary sensors into a single ground-level pass: RGB-D sensing for real-time citrus tree inventory with live field dashboards; hyperspectral canopy imaging for nutrient concentration estimation; proximal NIR soil spectroscopy for under-canopy soil reflectance signatures; and RTK-GNSS with onboard edge computing to georeferenced observations and deliver same-day data products without cloud connectivity. Results are validated against field surveys and ground-truth observations to ensure reliability for operational use.

Expected outcomes

• A real-time citrus tree inventory system validated at over 90% counting accuracy, with live results accessible from any device on the field network. No cloud or post-processing required.

• Georeferenced canopy nutrient maps (N, P, K, chlorophyll) derived from onboard hyperspectral inference, replacing multi-day laboratory turnaround with same-day GeoJSON outputs for prescription fertilization.

• Proximal soil spectral signatures at systematic within-row and midrow positions, supporting organic carbon and pH estimation across the soil profile.

• A validated multi-sensor robotic stack that completes canopy, soil, and tree count data collection in a single autonomous pass.

• Evaluation metrics matching the proposal targets: tree detection mean IoU ≥ 0.9, nutrient estimation R² ≥ 0.8, and soil property prediction R² ≥ 0.7, all delivered within the same field session.

Impact

By lowering technical barriers and accelerating in-field sensing, this project enables faster, more targeted agronomic decisions and more resilient agricultural management at scale. This project delivers same-day tree inventories, canopy nutrient status, and soil health information to growers and extension partners without complex GIS workflows or delayed laboratory turnaround.