Brandsma, Abe (2020) Analysis of Shape Classification Techniques in Hydrodynamic Imaging with an Artificial Lateral Line. Bachelor's Thesis, Artificial Intelligence.
|
Text
AI_BA_2020_ABRANDSMA.pdf Download (503kB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (93kB) |
Abstract
The lateral line is a unique sensing organ found in fish. It has been used as a model to create what is called an Artificial Lateral Line (ALL). This ALL can be used in hydrodynamic imaging for the classification of objects. This sensor array could replace traditional sensors. This research analyses the performance of two feature extraction techniques on two different flow datasets that have been created by the ALL. The difference of the flow datasets are characterized by two properties; the size of the sensor array and the size of the water volume in which the sensor array was placed. Two different feature extraction methods have been used; hand-picked feature extraction and an auto-encoder that automatically tries to detect which features to extract. These extracted features are then put through a classification algorithm called the Extreme Learning Machine. We found that the differences between the classification scores of the two flow datasets were small. However, there was a significant difference between the classification scores of individual windows. The auto-encoder resulted in non-generalizable features and low classification scores. The feature extraction methods mentioned in this paper are promising and could provide a fundamental enhancement of human sensing of underwater environments.
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Supervisor name: | Netten, S.M. van |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 03 Jul 2020 11:24 |
| Last Modified: | 03 Jul 2020 11:24 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/22365 |
Actions (login required)
 |
View Item |