Eirini Ntafi, "Reconfigurable logic-based real-time automatic video calibration and processing to detect holes in aquaculture nets", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2022
https://doi.org/10.26233/heallink.tuc.94013
As Computer Vision technology is developing more and more in the last decade, the areas of application expand to many research sectors, some of which are for underwater applications, such as in aquaculture. The main goal of this thesis was to create an embedded system that detects defective holes in aquaculture nets from underwater videos in various realistic conditions. First, we evaluate an embedded system that is already build against new video data that were taken in more realistic conditions, including marine fouling on the nets and fish in the background; such conditions were never tested before. We created a classification system according to the specific characteristics and problems that we identified as common in the videos. Subsequently we improve the detection capabilities of the system, either by using parameter calibration, or by adding an extra logic block in the existing system. After testing, the new system resulted on average x6 times better results over a broad range of realistic videos, which with the previous system were not detected at all. The software application was developed with MATLAB and the hardware design block was created using Vitis Unified Platform software (targeting the Alveo U50 Card). In conclusion, the system was improved to work on more realistic underwater conditions and addresses some of the limitations of the previous system.