Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

NIR hyperspectral imaging with machine learning to detect and classify codling moth infestation in apples

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  2021 ASABE Annual International Virtual Meeting  2100066.(doi:10.13031/aim.202100066)
Authors:   Nader Ekramirad, Alfadhl Y. Khaled, Lauren E. Doyle, Chadwick A. Parrish, Raul T. Villanueva, Kevin D. Donohue, Akinbode A. Adedeji
Keywords:   Apples, Codling moth, Hyperspectral imaging, Machine learning, Near-infrared

Abstract. Codling moth (CM) is one of the most devastating insect pests of apples in North America. Effective and early detection of external and internal CM infestation could remarkably reduce postharvest losses and improve the quality of local and exported apples. Hyperspectral imaging (HSI) has been used as a powerful tool for nondestructive defect detection and classification in agricultural products, with the advantage of providing both spectral and spatial features, and the ability to detect internal defects. These merits make HSI a suitable candidate for detecting CM infestation in apples, where the damage is mostly internal, occasionally with some hard-to-visualize surface symptoms such as holes and frass. In this study, a spectral-spatial classification method was used to distinguish CM-infested from non-infested apples based on near-infrared hyperspectral imaging (NIR HIS) in the wavelength range from 900 to 1700 nm with a 3.35 nm increment. Two approaches were applied and compared in this study. In the first approach mean reflectance spectra (MRS) were calculated for the image of the entire apple and the classification was performed with an overall test set classification rate of 81.04%. In the second approach, the fruit pixels were classified into two classes, control and infested, with a 99.24% total accuracy for the test data set from random forest (RF) classifier among others. These results indicate the high potential of NIR HSI in detecting and classifying CM infestation in apples.

(Download PDF)    (Export to EndNotes)