Automated shark detection using computer vision

Byles, K. (2016) Automated shark detection using computer vision. [USQ Project]

[img]
Preview
Text (Main Project)
Byles_K_Low.pdf

Download (17Mb) | Preview

Abstract

With the technological advancements of UAVs, researchers are finding more ways to harness their capabilities to reduce expenses in everyday society. Machine vision is at the forefront of this research and in particular image recognition. Training a machine to identify objects and di↵erentiate them from others plays an integral role in the advancement of artificial intelligence. This project aims to design an algorithm capable of automatically detecting sharks from a UAV. Testing is performed by post-processing aerial footage of sharks taken from helicopters and drones, and analysing the reliability of the algorithm.

Initially this research project involved analysing aerial photography of sharks, dissecting the images into the individual colour channels that made up the RGB and HSV colour spaces and identifying methods to detect the shark blobs. Once an adaptive threshold of the brightness channel was designed, filters were curated specific to the environments presented in the obtained aerial footage to reject false positives. These methods were considerably successful in both rejecting false positives and consistently detecting the sharks in the video feed.

The methods produced in this dissertation leave room for future work in the shark detection field. By acquiring more reliable data, improvements such as using a kalman filter to detect and track moving blobs could be implemented to produce a robust shark detection and tracking system.


Statistics for USQ ePrint 31382
Statistics for this ePrint Item
Item Type: USQ Project
Item Status: Live Archive
Additional Information: Bachelor of Engineering (Honours) Major Mechatronic Engineering project
Faculty / Department / School: Current - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering
Supervisors: Low, Tobias
Date Deposited: 19 Jul 2017 01:50
Last Modified: 20 Jul 2017 00:10
Uncontrolled Keywords: computer vision; automated shark dectection
Fields of Research : 09 Engineering > 0913 Mechanical Engineering > 091302 Automation and Control Engineering
URI: http://eprints.usq.edu.au/id/eprint/31382

Actions (login required)

View Item Archive Repository Staff Only