Sorting for internal flesh browning in apple using visible-shortwave near infrared spectroscopy

Fruit may develop internal disorders during growth and development or in storage. Typically, retailers set a specification of <2% of fruit affected by such disorders for a consignment to be accepted. Visible – short wave near infrared spectroscopy (Vis-SWNIR) has been suggested for detection of internal defects such as internal flesh browning in pome fruits like apple and pear. This paper documents online detection of apple (Pink Lady) internal flesh browning. A range of reference methods were explored, including a visual browning score (1-5 or 1-10 scale), image analysis of defect area of a cut fruit surface, color space (Lab) readings of the cut surface and absorbance of a juice extract were considered, with use of the visual 1-5 scale recommended. A range of optical geometries (interactance, partial and full transmission) were employed in both static and dynamic modes of operation. Fruit spectra were acquired using MMS1 and Avantes spectrometers and pre-processed using averaging, multiple scatter correction or standard normal variate. Wavelength ratio, partial least squares regression and classification tools were trialed, with multivariate analysis undertaken using Unscrambler 10.3 (Camo, Norway) and Matlab (Mathworks Inc.). Prediction statistics for PLSR models of around (r2p= 0.75, RMSEP= 0.21 were achieved. Sorting of defect from acceptable fruit was trialled using K Nearest neighbours (KNN), Partial Least Square Discriminant Analysis (PLS-DA), Linear Discriminant Analysis (LDA), Support Vector Machine (SVM) and multinomial logistic regression methods. The best results were obtained by using PLS-DA and multinomial logistic regression, with less than 5% of classification error (misclassification of good fruit as defect or vice versa). Wavelength ratio based method provided threshold adjustment opportunity to yield desirable result however losses as much as 20% of good fruit predicted defect to ensure less than 2% of defect fruit in good bin. Sorting of commercial lots of apple fruit with internal browning developed during controlled atmosphere storage using commercially available on-line equipment is demonstrated.