Mixture Regression for Sea Ice Segmentation
The classification of sea ice in SAR imagery is complicated by statistical nonstationarity. Incidence angle effects, heterogeneous ice conditions and other confounding variables contribute to spatial and temporal variability in the appearance of sea ice. I explore a family of models called mixture r...
Main Author: | |
---|---|
Format: | Master Thesis |
Language: | English |
Published: |
University of Waterloo
2022
|
Subjects: | |
Online Access: | http://hdl.handle.net/10012/19004 |
id |
ftunivwaterloo:oai:uwspace.uwaterloo.ca:10012/19004 |
---|---|
record_format |
openpolar |
spelling |
ftunivwaterloo:oai:uwspace.uwaterloo.ca:10012/19004 2023-05-15T15:40:36+02:00 Mixture Regression for Sea Ice Segmentation Manning, Max 2022-12-11 http://hdl.handle.net/10012/19004 en eng University of Waterloo http://hdl.handle.net/10012/19004 sea ice image segmentation nonstationary imagery mixture models mixture regression Master Thesis 2022 ftunivwaterloo 2022-12-24T23:57:48Z The classification of sea ice in SAR imagery is complicated by statistical nonstationarity. Incidence angle effects, heterogeneous ice conditions and other confounding variables contribute to spatial and temporal variability in the appearance of sea ice. I explore a family of models called mixture regressions which address this issue by endowing mixture distributions with class-dependent trends. I introduce mixture regression as a general technique for unsupervised clustering on nonstationary datasets and propose techniques to improve its robustness in the presence of noise and outliers. I then develop region-based mixture regression models for sea ice segmentation, focusing on the modeling of SAR backscatter intensities under the influence of incidence angle effects. Experiments are conducted on various extensions to the approach including the use of robust estimation to improve model convergence, the incorporation of Markov random fields for contextual smoothing, and the combination of mixture regression with supervised classifiers. Performance is evaluated for ice-water classification on a set of dual-polarized RADARSAT-2 images taken over the Beaufort Sea. Results show that mixture regression achieves accuracy of 92.8% in the unsupervised setting and 97.5% when integrated with a supervised convolutional neural network. This work improves on existing techniques for sea ice segmentation which enable operational ice mapping and environmental monitoring applications. The presented techniques may also be useful for the segmentation of nonstationary images obtained from other remote sensing techniques or in other domains such as medical imaging. Master Thesis Beaufort Sea Sea ice University of Waterloo, Canada: Institutional Repository |
institution |
Open Polar |
collection |
University of Waterloo, Canada: Institutional Repository |
op_collection_id |
ftunivwaterloo |
language |
English |
topic |
sea ice image segmentation nonstationary imagery mixture models mixture regression |
spellingShingle |
sea ice image segmentation nonstationary imagery mixture models mixture regression Manning, Max Mixture Regression for Sea Ice Segmentation |
topic_facet |
sea ice image segmentation nonstationary imagery mixture models mixture regression |
description |
The classification of sea ice in SAR imagery is complicated by statistical nonstationarity. Incidence angle effects, heterogeneous ice conditions and other confounding variables contribute to spatial and temporal variability in the appearance of sea ice. I explore a family of models called mixture regressions which address this issue by endowing mixture distributions with class-dependent trends. I introduce mixture regression as a general technique for unsupervised clustering on nonstationary datasets and propose techniques to improve its robustness in the presence of noise and outliers. I then develop region-based mixture regression models for sea ice segmentation, focusing on the modeling of SAR backscatter intensities under the influence of incidence angle effects. Experiments are conducted on various extensions to the approach including the use of robust estimation to improve model convergence, the incorporation of Markov random fields for contextual smoothing, and the combination of mixture regression with supervised classifiers. Performance is evaluated for ice-water classification on a set of dual-polarized RADARSAT-2 images taken over the Beaufort Sea. Results show that mixture regression achieves accuracy of 92.8% in the unsupervised setting and 97.5% when integrated with a supervised convolutional neural network. This work improves on existing techniques for sea ice segmentation which enable operational ice mapping and environmental monitoring applications. The presented techniques may also be useful for the segmentation of nonstationary images obtained from other remote sensing techniques or in other domains such as medical imaging. |
format |
Master Thesis |
author |
Manning, Max |
author_facet |
Manning, Max |
author_sort |
Manning, Max |
title |
Mixture Regression for Sea Ice Segmentation |
title_short |
Mixture Regression for Sea Ice Segmentation |
title_full |
Mixture Regression for Sea Ice Segmentation |
title_fullStr |
Mixture Regression for Sea Ice Segmentation |
title_full_unstemmed |
Mixture Regression for Sea Ice Segmentation |
title_sort |
mixture regression for sea ice segmentation |
publisher |
University of Waterloo |
publishDate |
2022 |
url |
http://hdl.handle.net/10012/19004 |
genre |
Beaufort Sea Sea ice |
genre_facet |
Beaufort Sea Sea ice |
op_relation |
http://hdl.handle.net/10012/19004 |
_version_ |
1766373189252808704 |