Using non-destructive laser backscattering imaging technology for kiwifruit quality assessment : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Food Technology at Massey University, Manawatū, New Zealand
Loading...
Files
Date
2024
DOI
Open Access Location
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Massey University
Figures 2.2, 2.5, 2.11, 4.1 and 4.2 are reproduced with permission.
Figures 2.2, 2.5, 2.11, 4.1 and 4.2 are reproduced with permission.
Rights
© The Author
Abstract
Kiwifruit is one of the most important exported horticultural products in New Zealand. The supply of kiwifruit to both national and international markets can be extended by harvesting kiwifruit unripe and storing with proper postharvest practice. During kiwifruit storage, quality monitoring is required for inventory planning and consistent quality maintenance. Currently, the industry is using sampled data to represent a batch of kiwifruit. However, kiwifruit quality is difficult to estimate based on destructively measured samples due to the heterogeneous population quality distribution. Therefore, a non-destructive technology is preferred allowing quality measurement for all kiwifruit prior to and during storage, as well as before exporting and marketing. Commercial spectral-optical devices, such as near-infrared (NIR) spectroscopy, have been employed by the industry for fruit grading and sorting at harvest, and have achieved good performance in total soluble solid content (SSC) and dry matter content (DMC) estimation. However, NIR spectroscopy had a poorer performance in estimating kiwifruit flesh firmness (FF), the primary quality indicator. During light and fruit tissue interaction, those optical devices capture data containing primarily the absorption signal related to kiwifruit’s chemical composition. Therefore, the FF estimation is indirect and the accuracy of FF measurement is affected when both textural structures and SSC change during postharvest ripening.
Laser backscattering imaging (LBI) records the backscattered signal after a single laser beam interacts with kiwifruit tissue. These light-tissue interactions include light absorption and scattering. The back-scattered signal could be analysed as an attenuation profile, and this attenuation profile is determined by optical properties of absorption (μa) and reduced scattering (μs’) coefficients, which correlate with fruit chemical compositions and physical properties, respectively. Therefore, LBI data is potentially helpful for FF estimation and early-stage internal disorder symptoms detection.
This PhD work developed a non-destructive approach based on the LBI technique to segregate kiwifruit with internal disorders [brown marmorated stink bug (BMSB) feeding injury and chilling injury (CI)], as well as soft fruit at FF threshold of 9.8 N. Estimation of μa and μs’ was achieved with 56.6 % and 91.5 % accuracy respectively, using a pre-classification method and validated against optical phantoms of known optical properties. Additionally, LBI parameters directly extracted from the images were utilised to develop segregation models owing to the uncertainties in μa and μs’ estimation.
For internal disorder detection, using the estimated kiwifruit μa and μs’, the segregation accuracy for kiwifruit with BMSB damage was 84 % and 62 % for ‘Zesy002’ (n=198) and ‘Hayward’ (n=198). Using extracted kiwifruit LBI parameters, the segregation accuracy for kiwifruit with CI was 92 % and 39 % for ‘Zesy002’ (n=396) and ‘Hayward’ (n=400). In addition, ‘Zesy002’ (n=30) and ‘Hayward’ (n=30) LBI during the postharvest ripening for kiwifruit were collected through a 15-day shelf life at 20 °C, where extracted LBI parameters were used to develop a time-series model. Absolute values of kiwifruit LBI parameters increased during the kiwifruit ripening process for both cultivars and the trend of LBI parameters may be correlated with kiwifruit softening. For segregating kiwifruit based on FF, the kiwifruit FF segregation model was calibrated, cross-validated and externally tested using kiwifruit LBI and corresponding FF data collected from 2 seasons with varying at-harvest maturity stages and stored at 2 temperatures. The segregation model accuracy for classifying fruit based on the 9.8 N FF threshold was 75 % and 70 % for ‘Zesy002’ (n=2247) and ‘Hayward’ (n=3558) in test sets. In conclusion, this work confirms that LBI technology has the potential for segregating soft kiwifruit or kiwifruit with early internal disorder symptoms and be adapted to the packhouse sorting system. However, in this work, FF segregation uncertainty at the 9.8 N threshold was observed when ‘Zesy002’ FF (N) ∈ (5,15) and ‘Hayward’ FF (N) ∈ (5,20) due to LBI parameter overlapping. Improved image analysis and segregation algorithms need to be investigated to enhance the segregation sensitivity for kiwifruit FF in the lower firmness range.
Description
Keywords
kiwifruit postharvest, non-destructive technology, time-series analysis, kiwifruit segregation, Kiwifruit, Postharvest technology, Postharvest physiology, Storage