Vis enkel innførsel

dc.contributor.advisorLillemo, Morten
dc.contributor.advisorMroz, Tomasz
dc.contributor.advisorShafiee, Sahameh
dc.contributor.authorLassegård, Henrik
dc.date.accessioned2022-03-18T14:37:16Z
dc.date.available2022-03-18T14:37:16Z
dc.date.issued2021
dc.identifier.urihttps://hdl.handle.net/11250/2986259
dc.description.abstractPhenotyping is a major bottleneck in breeding programs or crop-related field experiments. Development of high throughput phenotyping (HTP) methodologies holds promise of mitigating this shortcoming and providing applications capable of non-destructive and rapid recording of accurate phenotypes at large scales. In this work, three field experiments consisting of 300, 24 and 16 spring wheat varieties, respectively, were planted at Vollebekk research station in field season 2021 and phenotyped with DJI Phantom 4 drone across the growing season. Two sets of images captured by an unmanned aerial vehicle (UAV) at nominal altitudes of 20 and 8 meter above ground were used to estimate plot heights and model heading status of the plots by texture properties of the images. The estimated traits are compared with manually collected ground truth measurements to see whether the traits can be accurately described. One set of images was captured from at a 75 degree angle and used for generating digital surface models (DSM). The second set of images was captured at nadir and used to investigate how texture properties of the images relate to the heading process of the plots. Digital surface models (DSM) produced by Pix4D were used to estimate a terrain model which is used to produce estimates of the heights of the wheat plots. The estimated plot heights were compared to manual plot height measurements to assess the accuracy of the estimates. The DSMs were also used to provide altitude values for three-dimensional models of the plot surfaces. A dataset of uniform images depicting surfaces of known plots at known times was created from drone images calibrated and undistorted by Pix4D. Grey level co-occurrence (GLCM) texture features were extracted from the dataset and a logistic regression model used to assess the features’ ability to discriminate heading status of the plots.en_US
dc.language.isoengen_US
dc.publisherNorwegian University of Life Sciences, Åsen_US
dc.rightsAttribution-NonCommercial-NoDerivatives 4.0 Internasjonal*
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/deed.no*
dc.titleWheat trait prediction using UAVen_US
dc.typeMaster thesisen_US
dc.description.localcodeM-BIASen_US


Tilhørende fil(er)

Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel

Attribution-NonCommercial-NoDerivatives 4.0 Internasjonal
Med mindre annet er angitt, så er denne innførselen lisensiert som Attribution-NonCommercial-NoDerivatives 4.0 Internasjonal