Vegetation Height Estimation Using a 77-80 GHz Millimeter-Wave Radar

Espinal-De-La-Cruz D.; Clemente-Arenas M.; Lezama-Calvo J.; Ramos-Sanchez S. · 2026 IEEE 2nd Latin American Conference on Antennas and Propagation, LACAP 2026 - Proceedings (2026) · DOI: 10.1109/lacap67012.2026.11471184

Resumen

This paper investigates a low-cost 77-80 GHz FMCW millimeter-wave radar workflow for short-range forest canopy-height estimation. Radar echoes collected along linear trajectories in front of real trees are processed with a standard FMCW chain (range FFT, log-magnitude, smoothing, and CFAR) to generate a point cloud and a simple ground profile. Unsupervised clustering (DBSCAN, OPTICS, and Mean Shift) is then used to isolate canopy returns and estimate per-tree height as the difference between the canopy envelope and the ground surface. Experiments on three trees (3.36 ~m-3.45 ~m) show that DBSCAN and OPTICS achieve errors of about 0.2 ~m-0.3 ~m for two trees, while Mean Shift tends to merge nearby crowns and yields a larger error (0.6 m). These results suggest that FMCW mmWave radar combined with clustering can provide canopy-height estimates within a few decimeters for local forest monitoring when airborne LiDAR is impractical.

Instituciones: Universidad Nacional Tecnológica de Lima Sur