Non-Destructive Estimation of Leaf Area Index(LAI) Using Image Data
dc.contributor.advisor | Noble, Scott | |
dc.contributor.committeeMember | Helgason, Warren | |
dc.contributor.committeeMember | Meda, Venkatesh | |
dc.creator | Sediqi, Tansu | |
dc.date.accessioned | 2024-08-27T15:17:52Z | |
dc.date.available | 2024-08-27T15:17:52Z | |
dc.date.copyright | 2024 | |
dc.date.created | 2024-07 | |
dc.date.issued | 2024-08-27 | |
dc.date.submitted | July 2024 | |
dc.date.updated | 2024-08-27T15:17:53Z | |
dc.description.abstract | This study investigated using image data to estimate the Leaf Area Index (LAI) for wheat (Triticale aestivum L.). The calculation of LAI involved the utilization of light interception, and the estimation process involves creating a model using metrics extracted from Near-Infrared (NIR) and visible light image data. Line quantum sensors were used to collect above- and below-canopy light measurements. Standard models were used to calculate LAI based on these measurements. To address the effects of the sun azimuth angle, corrections were applied to the LAI values, effectively eliminating its impact on the determinations. The assessment of canopy coverage used in this study involved NIR and visible light images. To estimate LAI non-destructively, linear regression models were constructed, incorporating a set of metrics from the mentioned imaging modalities and the date as a proxy for the maturity of the wheat canopy. The performance of these models was evaluated using root mean squared error (RMSE) and adjusted coefficient of determination (R2). Through a comparison of these models, it was concluded that LAI was best estimated using NIR images, rather than visible-light images. Moreover, the most effective model for the estimation of LAI included only canopy coverage and maturity proxy as input parameters. Height was not an important factor in this case. The dataset examined contained a set of wheat varieties. To improve the power of LAI estimation models, the study employed a clustering technique. Rather than using a single, generic model for LAI estimation across all varieties, the proposed clustering strategy involves clustering the dataset and developing a model for each distinct cluster. This departure from the one-size-fits-all approach generated more precise LAI estimations for each cluster, ultimately enhancing the overall predictive accuracy of the models. This not only enhanced comprehension of LAI dynamics but also underscored the significance of personalized strategies in agricultural research to achieve more precise and applicable results. However, there are certain limitations to consider. The clusters showed inconsistencies, indicating possible difficulties in putting them into practice. While the suggestion of clustering exists, the specific approach to resolving these inconsistencies remains uncertain. | |
dc.format.mimetype | application/pdf | |
dc.identifier.uri | https://hdl.handle.net/10388/15959 | |
dc.language.iso | en | |
dc.subject | Leaf Area Index (LAI) | |
dc.subject | Light Interception | |
dc.subject | Line Quantum Sensors | |
dc.subject | Sun Azimuth Angle | |
dc.subject | Canopy Coverage | |
dc.subject | Linear Regression Models | |
dc.title | Non-Destructive Estimation of Leaf Area Index(LAI) Using Image Data | |
dc.type | Thesis | |
dc.type.material | text | |
thesis.degree.department | Mechanical Engineering | |
thesis.degree.discipline | Mechanical Engineering | |
thesis.degree.grantor | University of Saskatchewan | |
thesis.degree.level | Masters | |
thesis.degree.name | Master of Science (M.Sc.) |