Using image capture techniques to provide a representative image of per-plot spike density, a yield component of spring wheat (Triticum aestivum)
| dc.contributor.committeeMember | Shirtliffe, Steve | |
| dc.contributor.committeeMember | Pozniak, Curtis | |
| dc.contributor.committeeMember | Buekert, Rosalind | |
| dc.contributor.committeeMember | Eramian, Mark | |
| dc.contributor.committeeMember | Sharbel, Timothy | |
| dc.contributor.committeeMember | Si, Bingcheng | |
| dc.creator | Josuttes, Anique C | |
| dc.date.accessioned | 2020-12-10T02:53:37Z | |
| dc.date.available | 2022-12-09T06:05:08Z | |
| dc.date.created | 2020-10 | |
| dc.date.issued | 2020-12-09 | |
| dc.date.submitted | October 2020 | |
| dc.date.updated | 2020-12-10T02:53:37Z | |
| dc.description.abstract | Spike population density is an extremely important contributor to grain yield in wheat. Currently spike population density is not normally measured in breeding programs. The task is extremely labor intensive and may be inaccurate as a result of subsampling. High throughput phenotyping can aid in widening the bottle neck of traditional phenotyping but, images captured utilizing high throughput methods need to be representative of the infield spike population density if image analysis is to be correct. Wheat plots have a large level of occlusion that inhibits detection of all wheat spikes in the plot through the captured images. Exposure and time of image capture were tested to improve the representation and level of occlusion of in-field spike population density in captured RGB images. To do this, sixteen diverse wheat varieties were seeded into a randomized complete block design that contained three replications. The total number of wheat spikes per plot were counted. Image acquisition took place utilizing an ATV (All Terrain Vehicle) that had an RGB camera mounted on a wooden movable arm. All of the visible wheat spikes in the captured RGB images, that showcased the plot of interest from a zenith view, were annotated. The number of annotated wheat spikes per plot was compared to the infield spike count per plot. Human detected spike population density counts from the images had RSME values to in field spike population density of 22.32, 15.09, 17.58, 17.77, and 19.05. Results indicated that both exposure and time of image capture affected the ability to detect wheat spike in the captured images. The interaction between genotype and the image acquisition technique did not significantly affect the ability to detect wheat spikes in the captured images. The insignificant interaction was positive for the development of the high throughput pipeline as it indicated that given the image acquisition techniques and genotypes tested, the technique did not need to change depending on the genotype being imaged. This thesis showcased that simple image acquisition techniques can affect the ability to detect wheat spikes as well as improve occlusion in captured RGB images. | |
| dc.format.mimetype | application/pdf | |
| dc.identifier.uri | http://hdl.handle.net/10388/13168 | |
| dc.subject | Phenotyping | |
| dc.subject | Wheat | |
| dc.subject | Spike Density | |
| dc.subject | Imaging | |
| dc.title | Using image capture techniques to provide a representative image of per-plot spike density, a yield component of spring wheat (Triticum aestivum) | |
| dc.type | Thesis | |
| dc.type.material | text | |
| local.embargo.terms | 2022-12-09 | |
| thesis.degree.department | Plant Sciences | |
| thesis.degree.discipline | Plant Sciences | |
| thesis.degree.grantor | University of Saskatchewan | |
| thesis.degree.level | Masters | |
| thesis.degree.name | Master of Science (M.Sc.) |