Use of multispectral images and chemometrics in tomato seed studies

Publikation: KonferencebidragPosterForskning

Dokumenter

During the production of tomato seeds, green tomatoes are normally discarded before seed extraction irrespective of their maturity stage. Studies indicate that seeds from green tomatoes may reach be able to reach full germination capacity. Thus the potential of multispectral imaging for non-destructive discrimination of seeds based on their germination capacity was investigated. A total of 840 seeds extracted from green and red tomatoes were divided into two sets; a training set and a test set consisting of 648 and 192 seeds respectively. Each set consisted of 96 seeds from green tomatoes. The multispectral images of the seeds were captured and normalized canonical discriminant analysis was used to analyse the images. Germination tests were performed and seeds that subsequently germinated were recorded as viable. The viable seeds were classified with 99% and 98% accuracy for the training and test set, respectively. Similarly, dead seeds were predicted with 98% of accuracy. Results also showed that 23 and 14 seeds from green tomatoes in the training and test sets respectively were viable, while only one viable seed in the test set was misclassified. The results indicate that green tomatoes might be mature enough to contain viable seeds. However, it still needs to be investigated how these seeds perform in normal growing conditions.
OriginalsprogEngelsk
Udgivelsesår14 jun. 2016
Antal sider1
StatusUdgivet - 14 jun. 2016
Begivenhed31st ISTA Congress, Tallinn, Estonia 2016: ISTA seed symposium - Original Sokos Viru Hotel conference and banquet centre, Tallinn, Estland
Varighed: 15 jun. 201517 jun. 2016
https://ista-tallinn2016.ee/

Konference

Konference31st ISTA Congress, Tallinn, Estonia 2016
LokationOriginal Sokos Viru Hotel conference and banquet centre
LandEstland
ByTallinn
Periode15/06/201517/06/2016
Internetadresse

Se relationer på Aarhus Universitet Citationsformater

Download-statistik

Ingen data tilgængelig

ID: 103161305