Abstract
dc:descriptionThis thesis investigates non-destructive subsurface pest detection using multispectral imaging as the primary sensing method and thermal imaging as a secondary verification method. The goal was to determine whether imaging and convolutional neural network (CNN)-based analysis could distinguish pest-present and pest-absent samples in product-like containers before pests were easily visible by ordinary inspection. A controlled imaging system was developed using a BaySpec OCI-M multispectral camera, 530 nm green LED illumination, diffused white light, a two-mirror off-axis parabolic relay, conveyor-based sample motion, and a FLIR thermal camera. Experiments used rice and mixed grass, soil, and leaf backgrounds to represent different levels of visual and spectral complexity. Multispectral cubes were masked and normalized so a compact three-dimensional CNN could focus on container-region pest contrast across eight spectral bands. Thermal images were analyzed with a TinyUNet-style model to provide supporting heatmap-based verification. Results showed strong multispectral pest-detection performance across both backgrounds, while thermal imaging provided useful confirmation when temperature contrast was present. A preliminary late-fusion model and portable scanning prototype were also developed for future integrated inspection. Overall, the findings support multispectral imaging as a promising method for non-destructive subsurface pest detection, with thermal imaging serving as a complementary verification source.
Degree
thesis:*- Grantor dc:publisher
- University of North Texas
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Covington, Ellis 1997-
- Contributors dc:contributor
-
- Bostanci, Huseyin
- Manzo, Maurizio
- Qandil, Hassan
- Siu, King Man
Subjects
dc:subject × 3Rights
dc:rights- Statement dc:rights
-
- Public
- Covington, Ellis 1997-
- Copyright
- Copyright is held by the author, unless otherwise noted. All rights Reserved.
- Language dc:language
- English
Identifiers
dc:identifier.*- Identifier
-
local-cont-no: submission_5080
ark: ark:/67531/metadc2652566