Institutional Repository
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Determination of motion models in fish school

Ntoko Aneza

Full record


URI: http://purl.tuc.gr/dl/dias/BB9C2FEC-C565-46EB-BB31-984AFBFC1697
Year 2018
Type of Item Diploma Work
License
Details
Appears in Collections

Summary

The objective of the current thesis is the implementation, testing and validation of a video-surveillance system in fish farms that can be used in the future to analyze fish school behavior or their individual behavior. Such an application proves to be extremely useful in guiding the feeding procedure and inspect the quality of growth of fish populations in the expanding industry of fish farming. A crucial issue that we faced was the image enhancement of underwater images in videos, as they are characterized of intense noise, sudden light changes and blurring environment. Afterwards, the basic steps of our implementation involved the appropriate selection and combination of algorithms for detection, tracking and trajectory modeling of fish movements. The use of the Kalman Filter was important for the prediction and correction of fish motion. We further processed the trajectories after their extraction based on important observations on real videos. Finally, for the behavior analysis of fish we modeled and grouped trajectories into normal and abnormal classes, based on the uniformity of the trajectory slopes. As for the advantages of our system, we could say that future improvement is needed, but there is good indication that it can lead to a fully automatic system in order to be used under realistic conditions for real-time monitoring and interventions.

Available Files

Services

Statistics