LIGHT-WEIGHT MULTISPECTRAL UAV SENSORS AND THEIR CAPABILITIES FOR PREDICTING GRAIN YIELD AND DETECTING PLANT DISEASES

Abstract

Abstract. In this paper we investigate the performance of new light-weight multispectral sensors for micro UAV and their application to selected tasks in agronomical research and agricultural practice. The investigations are based on a series of flight campaigns in 2014 and 2015 covering a number of agronomical test sites with experiments on rape, barley, onion, potato and other crops. In our sensor comparison we included a high-end multispectral multiSPEC 4C camera with bandpass colour filters and reference channel in zenith direction and a low-cost, consumer-grade Canon S110 NIR camera with Bayer pattern colour filters. Ground-based reference measurements were obtained using a terrestrial hyperspectral field spectrometer. The investigations show that measurements with the high-end system consistently match very well with ground-based field spectrometer measurements with a mean deviation of just 0.01-0.04 NDVI values. The low-cost system, while delivering better spatial resolutions, expressed significant biases. The sensors were subsequently used to address selected agronomical questions. These included crop yield estimation in rape and barley and plant disease detection in potato and onion cultivations. High levels of correlation between different vegetation indices and reference yield measurements were obtained for rape and barley. In case of barley, the NDRE index shows an average correlation of 87% with reference yield, when species are taken into account. With high geometric resolutions and respective GSDs of down to 2.5 cm the effects of a thrips infestation in onion could be analysed and potato blight was successfully detected at an early stage of infestation.

Plant phenotyping relevance

軽量マルチスペクトルUAVセンサーの性能比較・地上計測との検証が中心で、収量推定や植物病害検出という植物表現型への適用も評価している。

Code and data availability

The paper describes UAV multispectral flight campaigns, NDVI/NDRE analyses, and yield/disease detection results, but contains no data availability statement, no public dataset deposit, and no author code release. The only URL present (www.blw.admin.ch) is a cited government reference, not a paper-specific asset.

No evidence-backed public reproduction asset is currently recorded.