A novel approach for Jatropha curcas seed health analysis based on multispectral and resonance imaging techniques

Abstract

Innovative methods have been developed in the state-of-the-art technologies based on robust spectral-spatial sensors for modern seed industry. In this study we proposed a novel approach based on multispectral imaging combined with machine learning algorithm to classify Jatropha curcas seed health. Furthermore, we present for the first time a methodology based on magnetic resonance imaging (MRI) to identify anatomical changes in J. curcas seeds infected with different pathogenic fungi. First, seeds were artificially inoculated with Lasiodiplodia theobromae, Colletotrichum siamense and Colletotrichum truncatum, and multispectral images were acquired after 24, 48, 72, 96, 120, 144 and 168 h of incubation. The MRI method was applied using incubated seeds for 168 h. Our results showed that the multispectral imaging technique combined with statistical models has the potential to distinguish different fungal species in J. curcas seeds after 48 h of incubation, with high accuracy (>80 %). The proposed MRI methodology allowed the identification of different damage patterns in the endosperm tissues infected with L. theobromae, C. siamense and C. truncatum. Therefore, multispectral imaging and MRI can be useful tools for rapid and accurate detection of different fungal species in J. curcas seeds.