Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Heroic failure or new dawn? Image-based identification of microalgae

Mann, David G., Droop, Stephen J. M., Hicks, Yulia Alexandrovna ORCID: https://orcid.org/0000-0002-7179-4587, Marshall, Andrew David ORCID: https://orcid.org/0000-0003-2789-1395, Martin, Ralph Robert and Rosin, Paul L. ORCID: https://orcid.org/0000-0002-4965-3884 2004. Heroic failure or new dawn? Image-based identification of microalgae. The Phycologist 66 , pp. 24-25.

Full text not available from this repository.

Abstract

A defensible estimate of species numbers in diatoms is c. 200,000, and evidence from several other microalgal groups indicate that current taxonomies are often ‘coarse-grained’ and do not discriminate sufficiently between biologically significant entities. However, making taxonomies more precise can make them difficult to use. Molecular methods offer one way around this problem and it has been suggested that sequence data could be used to ‘bar-code’ species for identification. However, this is currently unrealistic for many microalgae, because of rarity, poor sampling, recalcitrance in culture, or difficulty in obtaining sequences; furthermore, diatoms need to be identified when dead, for palaeoecology. The ADIAC and DIADIST projects were developed to make better use of morphological information, extracted without supervision from digital images, in classification and identification. Some new shape descriptors are highly sensitive and appear to surpass human visual capacity. DIADIST emulates traditional drawing in that a complex image is reduced to quantified, diagnostic essentials, which are then used for matching against a database of digitized drawings or photographs. DIADIST methods being developed can detect and represent striation patternsencouragingly well. Successful identification rates of > 95% in tests of ADIAC algorithms compare favourably with those achieved by experts. [EU, BBSRC funding].

Item Type: Article
Date Type: Publication
Status: Published
Schools: Computer Science & Informatics
Engineering
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Publisher: British Phycologist Society
ISSN: 0965-5301
Related URLs:
Last Modified: 10 Mar 2023 07:18
URI: https://orca.cardiff.ac.uk/id/eprint/44825

Actions (repository staff only)

Edit Item Edit Item