UAV3DCrop: Benchmarking 3D Reconstruction in Repeated Multi-Angle UAV Crop Surveys

2026-08-02

AI Summary

Existing 3D reconstruction methods struggle to simultaneously achieve photorealistic appearance, geometric accuracy, and agronomic utility in crop scenes, and lack standardized evaluation benchmarks tailored to repetitive multi-view drone imagery. This work introduces the first public benchmark dataset for precision agriculture, comprising 91 field plots of maize, soybean, wheat, and oat, along with 88,830 high-resolution RGB images. Two evaluation tracks are established: one focusing on optimized scene reconstruction using NeRF and 3D Gaussian Splatting (3DGS), and the other assessing zero-shot geometry estimation by pre-trained feedforward models such as MapAnything. Experiments show that Splatfacto-big achieves the best visual fidelity, while Scaffold-GS excels in depth and canopy height recovery; notably, only MapAnything reliably recovers absolute scale, whereas other feedforward models exhibit significant scale bias.

Abstract

Accurate 3D crop monitoring underpins data-driven precision agriculture by enabling field-scale analysis of plant structure, growth dynamics, and management response. Modern 3D reconstruction methods perform strongly on generic benchmarks, but rendered appearance may not translate into metrically and agronomically useful geometry in crop fields. We introduce UAV3DCrop, a public benchmark of repeated multi-angle unmanned aerial vehicle (UAV) crop surveys. It contains 88,830 RGB images at $5280 \times 3956$ pixels, with a ground sampling distance of 3.6-5.8 mm, from 91 scenes spanning corn, soybean, wheat, and oat. Track A evaluates seven scene-optimized methods -- Neural Radiance Field (NeRF) and 3D Gaussian Splatting (3DGS) variants -- on held-out views, photogrammetry-referenced depth, and canopy-height recovery. Track B tests four pretrained feed-forward models on zero-shot camera-pose and geometry estimation. The scene-optimized methods rank differently across the three targets: Splatfacto-big leads appearance, whereas Scaffold-GS leads depth and is statistically tied with Splatfacto for canopy height. Among feed-forward models, MapAnything leads on seven of the eight metrics, while the remaining models vary more across crops and fail severely on absolute scale in a way that alignment conceals. Repeated acquisitions reveal further sensitivities that differ by output type and by model, associated with position within the acquisition sequence and with tie-point multiplicity. Current 3D reconstruction methods are therefore not yet interchangeable for agronomic use: no single method wins on appearance, geometry, and canopy height at once, and only one of four feed-forward models recovers usable metric scale. The dataset is publicly available at https://link-dev.github.io/UAV3DCrop/

Problem

• 3D reconstruction
• precision agriculture
• UAV crop surveys
• canopy height
• metric scale

Innovation

• UAV3DCrop
• 3D reconstruction
• precision agriculture
• Neural Radiance Fields
• 3D Gaussian Splatting