Hyperspectral benthic mapping from underwater robotic platforms
Access status:
Open Access
Type
ThesisThesis type
Doctor of PhilosophyAuthor/s
Bongiorno, Daniel LucianoAbstract
We live on a planet of vast oceans; 70% of the Earth's surface is covered in water. They are integral to supporting life, providing 99% of the inhabitable space on Earth. Our oceans and the habitats within them are under threat due to a variety of factors. To understand the impacts ...
See moreWe live on a planet of vast oceans; 70% of the Earth's surface is covered in water. They are integral to supporting life, providing 99% of the inhabitable space on Earth. Our oceans and the habitats within them are under threat due to a variety of factors. To understand the impacts and possible solutions, the monitoring of marine habitats is critically important. Optical imaging as a method for monitoring can provide a vast array of information however imaging through water is complex. To compensate for the selective attenuation of light in water, this thesis presents a novel light propagation model and illustrates how it can improve optical imaging performance. An in-situ hyperspectral system is designed which comprised of two upward looking spectrometers at different positions in the water column. The downwelling light in the water column is continuously sampled by the system which allows for the generation of a dynamic water model. In addition to the two upward looking spectrometers the in-situ system contains an imaging module which can be used for imaging of the seafloor. It consists of a hyperspectral sensor and a trichromatic stereo camera. New calibration methods are presented for the spatial and spectral co-registration of the two optical sensors. The water model is used to create image data which is invariant to the changing optical properties of the water and changing environmental conditions. In this thesis the in-situ optical system is mounted onboard an Autonomous Underwater Vehicle. Data from the imaging module is also used to classify seafloor materials. The classified seafloor patches are integrated into a high resolution 3D benthic map of the surveyed site. Given the limited imaging resolution of the hyperspectral sensor used in this work, a new method is also presented that uses information from the co-registered colour images to inform a new spectral unmixing method to resolve subpixel materials.
See less
See moreWe live on a planet of vast oceans; 70% of the Earth's surface is covered in water. They are integral to supporting life, providing 99% of the inhabitable space on Earth. Our oceans and the habitats within them are under threat due to a variety of factors. To understand the impacts and possible solutions, the monitoring of marine habitats is critically important. Optical imaging as a method for monitoring can provide a vast array of information however imaging through water is complex. To compensate for the selective attenuation of light in water, this thesis presents a novel light propagation model and illustrates how it can improve optical imaging performance. An in-situ hyperspectral system is designed which comprised of two upward looking spectrometers at different positions in the water column. The downwelling light in the water column is continuously sampled by the system which allows for the generation of a dynamic water model. In addition to the two upward looking spectrometers the in-situ system contains an imaging module which can be used for imaging of the seafloor. It consists of a hyperspectral sensor and a trichromatic stereo camera. New calibration methods are presented for the spatial and spectral co-registration of the two optical sensors. The water model is used to create image data which is invariant to the changing optical properties of the water and changing environmental conditions. In this thesis the in-situ optical system is mounted onboard an Autonomous Underwater Vehicle. Data from the imaging module is also used to classify seafloor materials. The classified seafloor patches are integrated into a high resolution 3D benthic map of the surveyed site. Given the limited imaging resolution of the hyperspectral sensor used in this work, a new method is also presented that uses information from the co-registered colour images to inform a new spectral unmixing method to resolve subpixel materials.
See less
Date
2014-09-30Faculty/School
Faculty of Engineering and Information Technologies, School of Aerospace, Mechanical and Mechatronic EngineeringDepartment, Discipline or Centre
Australian Centre for Field RoboticsAwarding institution
The University of SydneyShare