Work

Četrti Doktorski dan Bi(o)znanosti?

Year: 2023
Type: book
Abstract: Crop breeding and phenotyping are crucial for developing improved crop varieties adapted to specific environments, increasing productivity, yield, and quality.This study aimed to test the applicability of a multispectral imaging system mounted on a UAV to identify 68 potato cultivars.Multispectral images were obtained using a 5-band multispectral camera mounted on a quadcopter at an altitude of 50 m AGL.Two datasets were built, and using each dataset, 40 spectral indices were calculated.The developed models achieved high success rates in identifying cultivars using per-pixel and mean spectra of plants.The per-pixel data had a success rate of 88.5%, while the mean spectra of plants had a success rate of 97.1%.The researchers combined per-pixel classifications with majority voting to achieve highly accurate results in classification tasks.These findings demonstrate the potential of spectral data and machine learning techniques in accurately identifying cultivars and enhancing crop breeding programs.Further studies, including time series, use of different environmental conditions, and combination with classical physiology assessment approaches are needed to validate these results and develop more generalizable models differentiation.
Authors: Mravlje, Jure, Starič, Pia
Language: Croatian
Subfield: Ecology
Open Access status: green
OpenAlex ID: W7036774443
FWCI:
Cites: 185
Cited by:
Related to: 10