Description
In the last decades, Halyomorpha halys emerged as a highly invasive pest causing serious damages to several agricultural crops making products unmarketable. Emilia-Romagna region was one of the areas of first occurrence of this pest in Europe. This region is also one of the most important for fruit production, accounting for 70% of pear cultivation in Italy. Since its discovery, H. halys caused severe economic losses to pear orchards, where in some cases more than 50% of harvested fruits presented damages due to punctures of this bug (Maistrello et al., 2017).
Near infrared hyperspectral imaging (NIR-HSI) can be suitable to identify damaged fruits during postharvest sorting to preserve pears quality. To this aim, hyperspectral images of punctured pears and control fruits were acquired in the 1156-1674 nm range at six subsequent times.
A preliminary image exploratory analysis step was performed using Principal Component Analysis (PCA) to point out the differences between sound and punctured areas. Nevertheless, the identification of Regions of Interest (ROIs) ascribable to the punctured regions resulted difficult, due to their strongly irregular shapes and blurred edges between sound and damaged areas.
To overcome this issue, a supervised annotation of punctured regions based on data dimensionality reduction methods was carried out, using the hyperspectrograms approach and spatial feature selection (Ferrari et al., 2013).
The conversion of the hyperspectral images into hyperspectrograms allowed to perform image-level classification between sound and punctured fruits, and to select relevant spatial features ascribable to the presence of punctures using interval Partial Least Squares Discriminant Analysis algorithm (iPLS-DA) (Ferrari et al., 2015). The features of interest were then visualized back into the original image domain, allowing an automatic selection of ROIs corresponding to punctured areas.
Keywords: Pests damage detection, pears, data dimensionality reduction, hyperspectral imaging, variable selection, multivariate classification
Acknowledgements: Study developed in the frame of HALY.ID project, which is part of ERA-NET Cofund ICT-AGRI-FOOD, with funding provided by national sources (Ministero delle politiche agricole e forestali, MIPAAF) and co-funding by the European Union’s Horizon 2020 research and innovation program, Grant Agreement number 862671.
REFERENCES
Ferrari, C., Foca, G., Calvini, R., Ulrici, A., 2015. Fast exploration and classification of large hyperspectral image datasets for early bruise detection on apples. Chemometrics and Intelligent Laboratory Systems 146, 108–119. https://doi.org/10.1016/j.chemolab.2015.05.016
Ferrari, C., Foca, G., Ulrici, A., 2013. Handling large datasets of hyperspectral images: Reducing data size without loss of useful information. Analytica Chimica Acta 802, 29–39. https://doi.org/10.1016/j.aca.2013.10.009
Maistrello, L., Vaccari, G., Caruso, S., Costi, E., Bortolini, S., Macavei, L., Foca, G., Ulrici, A., Bortolotti, P.P., Nannini, R., Casoli, L., Fornaciari, M., Mazzoli, G.L., Dioli, P., 2017. Monitoring of the invasive Halyomorpha halys, a new key pest of fruit orchards in northern Italy. Journal of Pest Science 90, 1231–1244. https://doi.org/10.1007/s10340-017-0896-2
