Estimation of chlorophyll content in banana during shelf life using LiDAR laser scanner

Abstract

Non-destructive sensing of fruit chlorophyll appears feasible using spectroscopy in the visible wavelength range. However, in postharvest application, measurements should even be carried out contact-free for avoiding disturbances of the produce and enabling automated measurements. In the present work a light detection and ranging sensor (LiDAR) laser scanner, emitting at 660 nm, was employed to analyse banana fruit (n = 180) with 0.9 m distance during fruit ripening for 8 d. The intensity of the echo was recorded at each laser hit, subsequently calibrated and geometrically corrected (Icor). With a grid size of 1 mm × 1 mm, mean Icor (n = 1200) of each fruit were further processed. The Icor and chemically analysed fruit chlorophyll content were used in univariate linear regression and partial least-square (PLS) regression analyses. The PLS approach resulted in enhanced R2 = 0.70, 0.63, and 0.77, with root mean squared error (RMSE) = 0.1808 mg cm-², 0.2007 mg cm-², and 0.0912 mg cm-² in the calibration, leave-one-out cross-validation, and test-set validation, respectively. Fruit classes were built applying Icor and either commercially used colour charts or chlorophyll-derived classes. Both classifications allowed to monitor the fruit in shelf life. The overall accuracy of classification with developed PLS-DA model were found 62 %, 61%, and 69 % for calibration, leave-one-out cross-validation, and test-set validation, respectively for the visually defined classes using colour chart, whereas 56 %, 53 %, and 66 % overall accuracies were achieved for the same on the true classes defined based on the chlorophyll content. Concluding, LiDAR laser scanning appears feasible for non-contact monitoring of banana fruit ripening.