Integration of drone imagery and artificial intelligence for high-throughput phenotypic selection of abiotic stress traits

Abstract

High-throughput phenotyping is a core prerequisite for breeding climate-resilient crops. To complete related breeding work, breeders must evaluate the performance of large-scale crop populations under seven types of field abiotic stresses including drought and high temperature, and the combined technology of unmanned aerial vehicle (UAV) imaging and artificial intelligence can provide core support to meet this demand. This review centers on three core sets of content: first, the integration of various UAV platforms, five types of imaging technologies, and machine learning and deep learning models to support phenotyping selection of abiotic stress-related traits; second, sorting out the biological significance of 12 categories of image-derived traits; third, breaking down the seven full workflow nodes ranging from flight planning to breeding decision support. Existing prior research on six crop types including wheat and rice has confirmed that this technology can improve the speed, scale and repeatability of field screening, and delivers outstanding effects when combined with multi-environment testing, genomic tools, and breeders’ expertise. This paper also sorts out six core limitations currently restricting the real-world deployment of this technology, and puts forward six future development directions to support its large-scale application.

Plant phenotyping relevance

UAV画像とAIによる作物のストレス関連形質の取得・選抜ワークフローを中心に整理したレビューであり、植物フェノタイピング手法が中核です。

Code and data availability

This is a narrative review of UAV-AI phenotyping for abiotic stress. It presents no paper-specific phenotype datasets, imagery, sensor data, analysis code, or trained models, and contains no availability or deposit statements for any such assets. All cited studies are prior work, not assets of this paper.

No evidence-backed public reproduction asset is currently recorded.