東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Linked to FindBook

Google Book

Amazon

博客來

Roadmap to Fully-Automated, Pan-Antarctic, Pack-Ice Seal Surveys.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Roadmap to Fully-Automated, Pan-Antarctic, Pack-Ice Seal Surveys./
Author:	Goncalves, Bento C.
Description:	1 online resource (131 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 84-06, Section: B.
Contained By:	Dissertations Abstracts International84-06B.
Subject:	Ecology. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30243241click for full text (PQDT)
ISBN:	9798363515309

Roadmap to Fully-Automated, Pan-Antarctic, Pack-Ice Seal Surveys.
Goncalves, Bento C.

Roadmap to Fully-Automated, Pan-Antarctic, Pack-Ice Seal Surveys. - 1 online resource (131 pages)

Source: Dissertations Abstracts International, Volume: 84-06, Section: B.

Thesis (Ph.D.)--State University of New York at Stony Brook, 2022.

Includes bibliographical references

Antarctic pack-ice seals, through their ecological role as key Antarctic krill predators, are critical to the Southern Ocean ecosystem. Shifts in sea ice distribution caused by anthropogenic climate change and krill fisheries threaten their populations. While initially surveyed by vessel or aircraft transects, very high-resolution remote sensing imagery has emerged as a safer and potentially cheaper alternative. The sheer volume of imagery, however, limits the spatial and temporal scale for human annotation of satellite imagery. AI-based, fully-automated surveys offer true scalability and, while imperfect, provide consistent annotations unaffected by observer fatigue or other factors external to the image itself. However, a pan-Antarctic survey using remote sensing comes with a number of challenges: 1) detecting seals in very-high-resolution imagery is a daunting task even for trained experts and relies heavily on contextual clues, making proper statistical treatment pivotal to go from detections to population estimates; 2) variability in lighting, terrain, off-nadir angle, and sea ice conditions impose severe limitations on the reliability of validation and test sets; and 3) limitations in our understanding of seal haul out behavior hamper our efforts to estimate the portion of seals available for detection (i.e. not submerged) at any moment in time. Here I present the recent advances in AI-powered seal detection and outline a schematic of the fully-automated pipeline that would be needed for regular pan-Antarctic seal surveying, along with requirements in terms of cost of imagery, personnel, and computational power. I will also discuss auxiliary components developed in support of an automated seal census, including sea ice segmentation models that are able to restrict input imagery to potential seal habitat only, human-level seal detection models, and the HPC middleware required to apply this efficiently at scale.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2023

Mode of access: World Wide Web

ISBN: 9798363515309Subjects--Topical Terms:

516476
Ecology.
Subjects--Index Terms:

AntarcticaIndex Terms--Genre/Form:

542853
Electronic books.

Roadmap to Fully-Automated, Pan-Antarctic, Pack-Ice Seal Surveys.
LDR:03409nmm a2200409K 4500 001 2354981
005 20230505090518.5
006 m o d
007 cr mn ---uuuuu
008 241011s2022 xx obm 000 0 eng d
020 $a 9798363515309
035 $a (MiAaPQ)AAI30243241
035 $a AAI30243241
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Goncalves, Bento C. $3 3695368
245 1 0 $a Roadmap to Fully-Automated, Pan-Antarctic, Pack-Ice Seal Surveys.
264 0 $c 2022
300 $a 1 online resource (131 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 84-06, Section: B.
500 $a Advisor: Lynch, Heather H.; Akcakaya, Resit R.
502 $a Thesis (Ph.D.)--State University of New York at Stony Brook, 2022.
504 $a Includes bibliographical references
520 $a Antarctic pack-ice seals, through their ecological role as key Antarctic krill predators, are critical to the Southern Ocean ecosystem. Shifts in sea ice distribution caused by anthropogenic climate change and krill fisheries threaten their populations. While initially surveyed by vessel or aircraft transects, very high-resolution remote sensing imagery has emerged as a safer and potentially cheaper alternative. The sheer volume of imagery, however, limits the spatial and temporal scale for human annotation of satellite imagery. AI-based, fully-automated surveys offer true scalability and, while imperfect, provide consistent annotations unaffected by observer fatigue or other factors external to the image itself. However, a pan-Antarctic survey using remote sensing comes with a number of challenges: 1) detecting seals in very-high-resolution imagery is a daunting task even for trained experts and relies heavily on contextual clues, making proper statistical treatment pivotal to go from detections to population estimates; 2) variability in lighting, terrain, off-nadir angle, and sea ice conditions impose severe limitations on the reliability of validation and test sets; and 3) limitations in our understanding of seal haul out behavior hamper our efforts to estimate the portion of seals available for detection (i.e. not submerged) at any moment in time. Here I present the recent advances in AI-powered seal detection and outline a schematic of the fully-automated pipeline that would be needed for regular pan-Antarctic seal surveying, along with requirements in terms of cost of imagery, personnel, and computational power. I will also discuss auxiliary components developed in support of an automated seal census, including sea ice segmentation models that are able to restrict input imagery to potential seal habitat only, human-level seal detection models, and the HPC middleware required to apply this efficiently at scale.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2023
538 $a Mode of access: World Wide Web
650 4 $a Ecology. $3 516476
650 4 $a Computer science. $3 523869
650 4 $a Remote sensing. $3 535394
653 $a Antarctica
653 $a Convolutional neural network
653 $a Deep learning
653 $a Pack-ice seals
653 $a Remote sensing
653 $a Very-high-resolution satellite imagery
655 7 $a Electronic books. $2 lcsh $3 542853
690 $a 0329
690 $a 0984
690 $a 0799
710 2 $a ProQuest Information and Learning Co. $3 783688
710 2 $a State University of New York at Stony Brook. $b Ecology and Evolution. $3 1675611
773 0 $t Dissertations Abstracts International $g 84-06B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30243241 $z click for full text (PQDT)