University of Groningen
Deep learning and hyperspectral imaging for unmanned aerial vehicles
Dijkstra, Klaas
DOI:
10.33612/diss.131754011
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Dijkstra, K. (2020). Deep learning and hyperspectral imaging for unmanned aerial vehicles: Combining convolutional neural networks with traditional computer vision paradigms. University of Groningen. https://doi.org/10.33612/diss.131754011
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Propositions accompanying the dissertation:
Deep Learning and Hyperspectral Imaging
for Unmanned Aerial Vehicles
Combining convolutional neural networks with traditional computer vision paradigms
by Klaas Dijkstra
1. Segmenting and counting many small and connected objects in images of bacterial colonies and images of potato-plant crops is best achieved by CentroidNet (this thesis, Chapter 5, pp. 126–136).
2. Combining deep learning with traditional computer-vision paradigms improves, compared to using only deep learning, the results for simultaneous localization and counting of potato-plant crops from aerial images (this thesis, Chapter 4, pp. 90–92).
3. The two tasks of hyperspectral demosaicking and crosstalk correction can be implemented by a single, dedicated three-layer convolutional neural network (this thesis, Chapter 3, pp. 66–70).
4. The seemingly unwanted effect of crosstalk between spectral bands in multi-color-filter-array sensors improves the performance of demosaicking (this thesis, Chapter 3, p. 63 and Chapter 6 pp. 142–143).
5. When reducing the number of spectral bands from 28 to 3, the majority of hyperspectral pixels of diseased-potato-plant leaves are still classified correctly using a k-nearest neighbor classifier (this thesis, Chapter 2, p. 19). 6. The trade offs in hyperspectral imaging can be understood by analyzing the various methods by which a 3-d hyperspectral cube is generated using a 2D imaging sensor (this thesis, Chapter 1, p. 6 and Chapter 3, p. 25).
7. Deep learning can be utilized to mitigate the resource limitations imposed by small aerial platforms that use (hyperspectral) imaging technology (this thesis, Chapter 6, pp. 141–143).
8. The future of deep learning and computer vision is in differentiable programming (this thesis, Chapter 6, p. 148).