Submitted:
16 November 2026
Posted:
18 November 2026
You are already at the latest version
Abstract
The rocoto (Capsicum pubescens) is a perennial shrub cultivated almost exclusively by small-scale Andean-Amazonian farmers on plots with scattered trees, where management is decided plant by plant. This study develops and validates a reproducible procedure for detecting individual plants and estimating their height from UAV photogrammetric point clouds. Two plots in Amazonas, Peru, of 6,758 and 15,010 m2, were surveyed through a complete census of 667 and 1,040 plants, whose heights were measured with a graduated ruler. The SfM-MVS point clouds were processed in R: ground classification with the cloth simulation filter, height normalisation, a canopy height model at 0.05 m, exclusion of woody vegetation by connected components, apex detection with a local maximum filter, and crown segmentation by height-constrained Voronoi. Detections were matched to the census one to one, with a tolerance set to half the planting spacing. A total of 533 and 1,105 crowns were delineated, with F-scores of 0.837 and 0.760. The mean canopy height within each crown was the most accurate height metric (R2 = 0.630 and 0.543; RMSE = 0.225 and 0.317 m), with systematic overestimation and compression of the range. Parameters calibrated on the first plot transferred to the second at a cost of 7.7 F-score points. False negatives concentrate near scattered trees, which defines the main condition for applicability.
Keywords:
UAV photogrammetry
; SfM-MVS
; canopy height model
; individual plant detection
; crop height
; high-throughput phenotyping
; smallholder agriculture
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.