ASABE Technical Library - Abstract

Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, view access and subscription options

HyperBird: A Hyperspectral Microscopic Imaging Robot for High Throughput Plant Phenotyping

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  2026 ASABE Annual International Meeting  2600738.(doi:10.13031/aim.202600738)
Authors:   Jinhong Yu, Aliyah Brewer, Lorenzo Pippi, Saeed Hosseinzadeh, Javier Moreno, Dani Martinez, Chang Chen, Kaitlin Gold, Lance Cadle-Davidson, Yu Jiang
Keywords:   Hyperspectral imaging; Microscopic imaging; High-throughput phenotyping; Plant disease detection; Grapevine downy mildew

Abstract. Hyperspectral imaging at microscopic resolution is a powerful engine for discovery in the plant sciences, especially when powered by AI-facilitated data analysis. However, slow data acquisition and the need for immense training datasets challenge system impact potential. We developed the HyperBird, an automated hyperspectral microscopic imaging robot designed for high-throughput data acquisition and scalable dataset generation to address these challenges. HyperBird integrates a push-broom hyperspectral camera with microscopic optics, a tungsten–halogen illumination module, and a Cartesian robotic arm within a fully automated imaging framework, enabling the acquisition of up to 351 sample wells in a single trial within three hours. System calibration established 24.3 µm (FWHM) spatial and 0.6 nm spectral resolutions respectively, producing 2000 x 2195 pixel by 950 narrow spectral bands (400–1000 nm) hyperspectral data cubes. Thermal evaluation confirmed the 150 W illumination setting causes minimal sample heating during scanning with no negative impact to disease development and progression. We present a case study with grapevine downy mildew (Plasmopara viticola) inoculated leaves continuously scanned over 0 to 6 days post inoculation (dpi) to illustrate the value of HyperBird's high spatial–spectral resolution. We built an automated pipeline to generate infected tissue masks based on 6 dpi extent and retroactively applied them to 0 through 5 dpi hypercubes to characterize disease progression-related spectral changes. Mixed-effects modeling revealed that spatially resolved spectral data yielded a 6.1-unit increase in progression slope (p < 0.001) and more consistent progression quantification compared to object-level spectral averaging. These results highlight HyperBird as a high-throughput platform for generating high-quality hyperspectral datasets, resolving spectral features through enhanced spatial resolution, and supporting more sensitive plant spectral analysis, thereby addressing key limitations in current sensing and AI-driven plant research workflows.

(Download PDF)    (Export to EndNotes)