Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping
Backscatter information from multibeam echosounders (MBES) have been shown to contain useful information for the characterisation of benthic habitats. Compared to backscatter imagery, angular response of backscatter has shown advantages for feature discrimination. However its low spatial resolution...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Published: |
Academic Press
2011
|
| Subjects: |
| _version_ | 1796858620052242432 |
|---|---|
| author | Che Hasan, Rozaimi Ierodiaconou, Daniel Laurenson, Laurie |
| author_facet | Che Hasan, Rozaimi Ierodiaconou, Daniel Laurenson, Laurie |
| author_sort | Che Hasan, Rozaimi |
| collection | ePrints |
| description | Backscatter information from multibeam echosounders (MBES) have been shown to contain useful information for the characterisation of benthic habitats. Compared to backscatter imagery, angular response of backscatter has shown advantages for feature discrimination. However its low spatial resolution inhibits the generation of fine scale habitat maps. In this study, angular backscatter response was combined with image segmentation of backscatter imagery to characterise benthic biological habitats in Discovery Bay Marine National Park, Victoria, Australia. Angular response of backscatter data from a Reson Seabat 8101 MBES (240 kHz) was integrated with georeferenced underwater video observations for constructing training data. To produce benthic habitat maps, decision tree supervised classification results were combined with mean shift image segmentation for class assignment. The results from mean angular response characteristics show effects of incidence angle at the outer angle for invertebrates (INV) and mixed red and invertebrates (MRI) classes, whilst mixed brown algae (MB) and mixed brown algae and invertebrates (MBI) showed similar responses independent from incidence angle. Automatic segmentation processing produce over segmented results but showed good discrimination between heterogeneous regions. Accuracy assessment from habitat maps produced overall accuracies of 79.6% (Kappa coefficient = 0.66) and 80.2% (Kappa coefficient = 0.67) for biota and substratum classifications respectively. MRI and MBI produced the lowest average accuracy while INV the highest. The ability to combine angular response and backscatter imagery provides an alternative approach for investigating biological information from acoustic backscatter data. |
| first_indexed | 2024-03-05T19:15:09Z |
| format | Article |
| id | utm.eprints-44790 |
| institution | Universiti Teknologi Malaysia - ePrints |
| last_indexed | 2024-03-05T19:15:09Z |
| publishDate | 2011 |
| publisher | Academic Press |
| record_format | dspace |
| spelling | utm.eprints-447902017-09-13T04:17:53Z http://eprints.utm.my/44790/ Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping Che Hasan, Rozaimi Ierodiaconou, Daniel Laurenson, Laurie TR Photography Backscatter information from multibeam echosounders (MBES) have been shown to contain useful information for the characterisation of benthic habitats. Compared to backscatter imagery, angular response of backscatter has shown advantages for feature discrimination. However its low spatial resolution inhibits the generation of fine scale habitat maps. In this study, angular backscatter response was combined with image segmentation of backscatter imagery to characterise benthic biological habitats in Discovery Bay Marine National Park, Victoria, Australia. Angular response of backscatter data from a Reson Seabat 8101 MBES (240 kHz) was integrated with georeferenced underwater video observations for constructing training data. To produce benthic habitat maps, decision tree supervised classification results were combined with mean shift image segmentation for class assignment. The results from mean angular response characteristics show effects of incidence angle at the outer angle for invertebrates (INV) and mixed red and invertebrates (MRI) classes, whilst mixed brown algae (MB) and mixed brown algae and invertebrates (MBI) showed similar responses independent from incidence angle. Automatic segmentation processing produce over segmented results but showed good discrimination between heterogeneous regions. Accuracy assessment from habitat maps produced overall accuracies of 79.6% (Kappa coefficient = 0.66) and 80.2% (Kappa coefficient = 0.67) for biota and substratum classifications respectively. MRI and MBI produced the lowest average accuracy while INV the highest. The ability to combine angular response and backscatter imagery provides an alternative approach for investigating biological information from acoustic backscatter data. Academic Press 2011 Article PeerReviewed Che Hasan, Rozaimi and Ierodiaconou, Daniel and Laurenson, Laurie (2011) Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping. Estuarine, Coastal and Shelf Science, 97 . pp. 1-9. ISSN 0272-7714 |
| spellingShingle | TR Photography Che Hasan, Rozaimi Ierodiaconou, Daniel Laurenson, Laurie Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping |
| title | Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping |
| title_full | Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping |
| title_fullStr | Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping |
| title_full_unstemmed | Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping |
| title_short | Combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping |
| title_sort | combining angular response classification and backscatter imagery segmentation for benthic biological habitat mapping |
| topic | TR Photography |
| work_keys_str_mv | AT chehasanrozaimi combiningangularresponseclassificationandbackscatterimagerysegmentationforbenthicbiologicalhabitatmapping AT ierodiaconoudaniel combiningangularresponseclassificationandbackscatterimagerysegmentationforbenthicbiologicalhabitatmapping AT laurensonlaurie combiningangularresponseclassificationandbackscatterimagerysegmentationforbenthicbiologicalhabitatmapping |