Multiple ortho‐mosaicking software pipelines produce comparable imagery‐derived wheat phenotypes

Abstract

Abstract Unmanned aerial systems (UAS) equipped with multispectral and RGB sensors offer valuable data for monitoring crop health and assessing disease severity. However, the wide range of available photogrammetric software complicates software selection for high‐throughput plant phenotyping. This study compares the consistency of three widely used software packages, OpenDroneMap (ODM), Agisoft Metashape, and Pix4Dmapper in processing UAS‐acquired imagery for wheat ( Triticum aestivum L.) phenotyping. Over two seasons at Oklahoma State University research sites, imagery from a DJI Phantom 4 Pro Multispectral was used to generate eight vegetation indices (VIs), which were evaluated using correlation () and root mean square error (RMSE). Normalized VIs showed strong consistency across software, with values between 0.85 and 0.99 and RMSEs ranging from 0.004 to 0.07. Non‐normalized indices exhibited greater variability but retained high correlations ( > 0.76). Ground‐truth validation used single‐view imagery and disease severity ratings. Bayesian models quantified spectral measurement differences, their distributions (mean, standard deviation, skewness, and excess kurtosis) across processing approaches, and evaluated disease classification performance using ordinal logistic regression. Normalized VIs were highly consistent across software (posterior median differences <0.01 units, overlapping 95% highest density intervals), while single‐view imagery showed 15%–25% higher pixel‐level variability than software outputs. Non‐normalized indices showed greater processing sensitivity. RGB indices demonstrated near to perfect consistency. Disease classification accuracy ranged from 35% to 48% with minimal software differences (<2%). All three software produce biologically consistent results, ensuring stable genotype rankings regardless of processing choice. ODM performed comparably to proprietary alternatives while offering cost‐effectiveness, transparency, and reproducibility advantages.

Plant phenotyping relevance

UAS画像から抽出する小麦表現型について、複数の写真測量ソフトウェアの一貫性・誤差・再現性を比較検証しており、フェノタイピング手法の技術評価が中心です。

Code and data availability

The article describes UAS wheat phenotyping comparing ODM, Metashape, and Pix4D, but no blocks contain author-deposited phenotype datasets, imagery, analysis code, or trained models with explicit public availability language. The only URLs (github.com/OpenDroneMap/ODM, agisoft.com, pix4d.com) are generic third-party软件,

No evidence-backed public reproduction asset is currently recorded.