Multispectral imaging as a potential tool for seed health testing of spinach (Spinacia oleracea L.)

Seed health tests are time consuming and require substantial training for characterization of pathogenic fungi on seed. A new approach to use a multispectral vision system for identifying surface properties of different fungal infections has been tested in spinach (Spinacia oleracea L.) at Aarhus University.

Our study demonstrates that multispectral imaging with wavelengths ranging from 395-970 nm can be used to distinguish between uninfected spinach seeds and seeds infected with Verticillium spp., Fusarium spp., Stemphylium botryosum, Cladosporium spp. and Alternaria alternata. Analytical separation based on mean pixel intensity, Canonical Discriminant Analysis (CDA) and classification by Jeffries-Matusita (JM) distance illustrates that a combination of Near Infrared spectra (NIR) and Visual spectra (VIS) is able to identify uninfected seeds from infected seeds ranging from 80-100%.

Classification using only NIR gave a separation of 26-88% between uninfected and Fusarium spp. infected seeds. Alternaria alternata and Fusarium spp. could be distinguished from each other and from Cladosporium spp., Verticillium spp. and Stemphylium spp. Separation of Cladosporium spp., Verticillium spp. and Stemphylium spp. needs further development before practical application.