Detecting Scallops in Images from an AUV
The University of Iceland owns an AUV and intends to use it for scallop abundance estimation. In November 2011 the AUV acquired images of the seabed in Breiðafjörður Fjord. These images are the basis for the topic of this thesis: seabed imaging and automatic detection of scallops with a scallop dete...
Main Author: | |
---|---|
Other Authors: | |
Format: | Thesis |
Language: | English |
Published: |
2012
|
Subjects: | |
Online Access: | http://hdl.handle.net/1946/13272 |
id |
ftskemman:oai:skemman.is:1946/13272 |
---|---|
record_format |
openpolar |
spelling |
ftskemman:oai:skemman.is:1946/13272 2023-05-15T15:46:30+02:00 Detecting Scallops in Images from an AUV Einar Óli Guðmundsson 1986- Háskóli Íslands 2012-10 application/pdf http://hdl.handle.net/1946/13272 en eng http://hdl.handle.net/1946/13272 Iðnaðarverkfræði Neðansjávarmyndir Fiskifræði Stofnstærð (vistfræði) Hörpudiskur Thesis Master's 2012 ftskemman 2022-12-11T06:56:52Z The University of Iceland owns an AUV and intends to use it for scallop abundance estimation. In November 2011 the AUV acquired images of the seabed in Breiðafjörður Fjord. These images are the basis for the topic of this thesis: seabed imaging and automatic detection of scallops with a scallop detector. The scallop detector first scans an image with a detection window and uses a classifier to predict wether a window includes a scallop or not. A voting method is then used on the positive windows to discard unwanted detections. First, the calibration of the AUV’s camera is conducted and the results show minimal lens distortion affecting the camera. The training set is made from the AUV’s images and then used for feature extraction. Two simple feature extraction methods are applied: gray-level thresholding; and color histograms. Three classification methods are proposed: nearest neighbor algorithm; distance to the average feature image; and SVMs. The classification methods are then used by the scallop detector. Experiments on the classifiers show that SVMs outperform the two other methods. Tuning of voting parameters is then conducted using three different scallop detectors. The combination detector, which first uses the color histogram classifier to find the most prominent area of the images and then the gray-level classifier on that area, shows the best results. Finally, the combination detector is tested on 20 images. The detector shows about 80% prediction accuracy. Háskóli Íslands á kafbát sem nota á til að meta stofnstærð hörpudisks. Í nóvember 2011 tók kafbáturinn myndir af hafsbotni Breiðafjarðar. Þessar myndir eru grunnurinn að umfjöllunarefni ritgerðarinnar; myndataka af hafsbotni og sjálfvirk talning hörpudiska með hörpudiskateljara. Hörpudiskateljarinn byrjar á því að skanna mynd með talningarglugga og notar svo flokkara til að spá fyrir um hvort glugginn innihaldi hörpudisk eða ekki. Kosningaraðferð er svo notuð á jákvæðu gluggana til að losna við óæskilegar talningar. Fyrst var kvörðun á myndavél ... Thesis Breiðafjörður Iceland Skemman (Iceland) Háskóli Íslands ENVELOPE(-21.949,-21.949,64.141,64.141) Breiðafjörður ENVELOPE(-23.219,-23.219,65.253,65.253) |
institution |
Open Polar |
collection |
Skemman (Iceland) |
op_collection_id |
ftskemman |
language |
English |
topic |
Iðnaðarverkfræði Neðansjávarmyndir Fiskifræði Stofnstærð (vistfræði) Hörpudiskur |
spellingShingle |
Iðnaðarverkfræði Neðansjávarmyndir Fiskifræði Stofnstærð (vistfræði) Hörpudiskur Einar Óli Guðmundsson 1986- Detecting Scallops in Images from an AUV |
topic_facet |
Iðnaðarverkfræði Neðansjávarmyndir Fiskifræði Stofnstærð (vistfræði) Hörpudiskur |
description |
The University of Iceland owns an AUV and intends to use it for scallop abundance estimation. In November 2011 the AUV acquired images of the seabed in Breiðafjörður Fjord. These images are the basis for the topic of this thesis: seabed imaging and automatic detection of scallops with a scallop detector. The scallop detector first scans an image with a detection window and uses a classifier to predict wether a window includes a scallop or not. A voting method is then used on the positive windows to discard unwanted detections. First, the calibration of the AUV’s camera is conducted and the results show minimal lens distortion affecting the camera. The training set is made from the AUV’s images and then used for feature extraction. Two simple feature extraction methods are applied: gray-level thresholding; and color histograms. Three classification methods are proposed: nearest neighbor algorithm; distance to the average feature image; and SVMs. The classification methods are then used by the scallop detector. Experiments on the classifiers show that SVMs outperform the two other methods. Tuning of voting parameters is then conducted using three different scallop detectors. The combination detector, which first uses the color histogram classifier to find the most prominent area of the images and then the gray-level classifier on that area, shows the best results. Finally, the combination detector is tested on 20 images. The detector shows about 80% prediction accuracy. Háskóli Íslands á kafbát sem nota á til að meta stofnstærð hörpudisks. Í nóvember 2011 tók kafbáturinn myndir af hafsbotni Breiðafjarðar. Þessar myndir eru grunnurinn að umfjöllunarefni ritgerðarinnar; myndataka af hafsbotni og sjálfvirk talning hörpudiska með hörpudiskateljara. Hörpudiskateljarinn byrjar á því að skanna mynd með talningarglugga og notar svo flokkara til að spá fyrir um hvort glugginn innihaldi hörpudisk eða ekki. Kosningaraðferð er svo notuð á jákvæðu gluggana til að losna við óæskilegar talningar. Fyrst var kvörðun á myndavél ... |
author2 |
Háskóli Íslands |
format |
Thesis |
author |
Einar Óli Guðmundsson 1986- |
author_facet |
Einar Óli Guðmundsson 1986- |
author_sort |
Einar Óli Guðmundsson 1986- |
title |
Detecting Scallops in Images from an AUV |
title_short |
Detecting Scallops in Images from an AUV |
title_full |
Detecting Scallops in Images from an AUV |
title_fullStr |
Detecting Scallops in Images from an AUV |
title_full_unstemmed |
Detecting Scallops in Images from an AUV |
title_sort |
detecting scallops in images from an auv |
publishDate |
2012 |
url |
http://hdl.handle.net/1946/13272 |
long_lat |
ENVELOPE(-21.949,-21.949,64.141,64.141) ENVELOPE(-23.219,-23.219,65.253,65.253) |
geographic |
Háskóli Íslands Breiðafjörður |
geographic_facet |
Háskóli Íslands Breiðafjörður |
genre |
Breiðafjörður Iceland |
genre_facet |
Breiðafjörður Iceland |
op_relation |
http://hdl.handle.net/1946/13272 |
_version_ |
1766381179762638848 |