Underwater Object Recognition with a Remotely OperatedVehicle
Author
Cehs, Vitalijs
Term
4. term
Publication year
2020
Pages
71
Abstract
Denne afhandling adresserer udfordringen med at identificere affald og andre objekter under vand som led i arbejdet med at reducere forurening (bl.a. relateret til FN’s Verdensmål 14). Projektets forskningsspørgsmål er: Hvordan udvikles et objektdetektionssystem, der kan udføre realtidsanalyse under vand for at finde og identificere relevante objekter? Arbejdet omfatter en gennemgang af state-of-the-art inden for undervandsfartøjer, maskinlæring og computer vision samt design af en systemarkitektur baseret på undervandsdronen BlueROV2. Prototypen kombinerer Python, TensorFlow, OpenCV og YOLOv3 til billedgenkendelse, og udviklingsprocessen følger en iterativ metode inspireret af Iterative Waterfall og Prototyping. Systemet implementeres på BlueROV2 og evalueres gennem iscenesatte eksperimenter under vand for at vurdere dets anvendelighed til realtidsdetektion. Kvantitative resultater eller præstationsmålinger er ikke beskrevet i det leverede uddrag.
This thesis addresses the challenge of identifying litter and other objects underwater as part of efforts to reduce pollution (related to UN Sustainable Development Goal 14). The core research question is: How can an object detection system be developed to perform real-time underwater analysis to detect and identify objects of interest? The work includes a review of the state of the art in underwater vehicles, machine learning, and computer vision, and proposes a system architecture built around the BlueROV2 underwater drone. The prototype integrates Python, TensorFlow, OpenCV, and the YOLOv3 model for visual recognition, and the development follows an iterative process inspired by the Iterative Waterfall and Prototyping methods. The system is implemented on the BlueROV2 and assessed through a series of staged underwater experiments to gauge its suitability for real-time detection. Quantitative outcomes or performance metrics are not provided in the supplied excerpt.
[This summary has been generated with the help of AI directly from the project (PDF)]
Keywords
Documents