Multispectral images for monitoring fruit ripeness: validation methods

Multi-spectral images have been previously used to classify different kinds of fruits into postharvest maturity classes. Previous research has succeeded in classifying peaches into ripeness clusters, gathering the whole variability of ripeness in the harvest and post-harvest chain (Lleó et al, 2007). The main objective of this work is to correlate fruit image with firmness and other quality traits through the analysis and classification of R-IR images on commercial conditions. MT-firmness and colour Reflectance parameters are used as reference values. The second objective is to develop an automatic procedure, able to classify on line commercial peaches into ripening classes consistently correlated with fruit quality. During the 2006 season, 500 images (2 images per fruit) including just harvested peaches, and over ripened fruits, in order to simulate commercial conditions, were considered for the generation of 6 ripeness classes. For external validation 1304 images from on season 2006; and 1020 images from season 2007 were analyzed. For both seasons, image-based classes showed constant trends and coherent ranges on their reference values: high percentage (91% for season 2006 and 81% for 2007) of the samples below minimum firmness values for transport (Crisosto, 1996) were classified into clusters 4, 5 or 6. The studied method shows a good potential to characterize the ripening state of the fruits, although further research is required to ensure high reliability of the system.