Login | Request Account (DAF staff only)

Hyperspectral imaging to non-destructively predict citrus chilling injury

Duong, H., Macnish, A. J., Wright, C. L. and Wedding, B. B. (2021) Hyperspectral imaging to non-destructively predict citrus chilling injury. Project Report. State of Queensland.

[img]
Preview
PDF
940kB

Abstract

Queensland lemons are exported at low (i.e. <3°C) temperature to meet market access protocols. While the fruit generally tolerate these conditions, skin damage consistent with chilling injury (CI) is occasionally encountered and results in significant economic and reputational loss. The robustness of fruit to withstand export handling is inherently variable and difficult to predict. This study evaluated the potential of hyperspectral imaging to detect and predict CI in fruit before it was visible to the human eye.
Seeded and seedless Eureka lemons were sourced from a north Queensland orchard. They were treated with either 2.6 or 18.5 parts per million (ppm) ethylene gas at 20.4 or 28.9oC for 5 days to hasten the loss of green skin colour, as per commercial degreening practice. The degreened fruit were stored at 1.9oC for 5 weeks to simulate seafreight to export markets, followed by 10 days at 20oC to assess CI under typical retail display conditions.
Spectral data was collected within 2-3 h after harvest, degreening or cold storage. Three near infrared (NIR) devices were used; Bruker Matrix-F for spot assessment; Pika NIR 320 and Pika XC2 for hyperspectral imaging, covering wavelengths from 830-2500 nm, 900-1700 nm and 400-1000 nm, respectively. The supervised classification method of linear discriminate analysis was applied to the spectra collected from the Bruker Matrix-F, whereas partial least squares discriminant analysis (PLS-DA) was used to analyse spectra and images obtained from the Pika XC2 and NIR 320 instruments.
The results showed that there is potential for NIR hyperspectral imaging to predict CI in seeded and seedless Eureka lemons, when assessed after degreening and particularly after cold shipment. Spectra collected after degreening and cold storage demonstrated reasonable levels (i.e. 70%) of correct classification. The highest correct classification was obtained for scans taken after storage and is likely due to CI having been induced in the fruit. Spectra collected 2-3 h after harvest using the Bruker Matrix-F device were not analysed due to fruit damage from the technique itself.
Future research is needed to explore wavelength selections to refine, verify and develop models against an independent prediction set that combines the two varieties. The maximum level of spatial and/or spectral binning for the hyperspectral images that results in minimal loss of information should also be investigated. Once models are tested and validated successfully at a laboratory level, there would be scope for RD&E investment in a model that could be integrated into handheld or in-line NIR sensors to assist industry in grading out sensitive fruit batches that lack robustness to withstand export.

Item Type:Monograph (Project Report)
Corporate Creators:Department of Agriculture and Fisheries
Business groups:Horticulture and Forestry Science
Keywords:Final report Agri-Science Queensland Innovation Opportunity
Subjects:Agriculture > Agriculture (General) > Agricultural economics
Plant culture > Harvesting, curing, storage
Plant culture > Food crops
Plant culture > Fruit and fruit culture
Deposited On:22 Sep 2021 02:09
Last Modified:22 Sep 2021 02:09

Repository Staff Only: item control page

Downloads

Downloads per month over past year

View more statistics