Institutional Repository of Vilnius University
Pollen recognition from a pair of holographic images /
Abstract
dc:descriptionThis study uses holographic imaging data from the Swisens Poleno Mars device to categorize holographic images into 13 pollen types, non-pollen particles, and water droplets, which addresses the limitations of standard pollen identification methods. A CNN-based framework was created, including sophisticated approaches like data augmentation, confidence thresholding, and feature extraction to improve accuracy, scalability, and resilience. The model underwent iterative testing, obtaining an overall accuracy of 88% and closely matching Hirst volumetric trap data. It outperformed standard methods and the proprietary Swisens AG model, especially in dominant pollen classes such as Alder, Birch, and Hazel, while efficiently decreasing false positives with stringent confidence thresholding and supplemented datasets. Computational optimization with Intel's OpenVINO toolbox improved inference speed and memory economy, making it suitable for real-time deployments. The model proved the ability to detect temporal pollen trends, providing a dependable and scalable automated pollen classification solution. By lowering manual effort and boosting accuracy, this study provides the framework for future advances in model architecture, addressing seasonal and geographic fluctuations, and expanding datasets to improve allergy predictions.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Senthil Kumar, Nikhil,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD229558665&prefLang=en_US