Development of new measurement methods to determine sugarcane quality from stalk samples

Development of new measurement methods to determine sugarcane quality from stalk samples

Recently, there has been a growing interest within the Australian sugarcane industry to measure sugarcane quality in the field to further improve product quality and value. However, conventional technologies for measuring sugarcane quality in a laboratory have limitations for uses in the field because they require sugarcane to be prepared as either juice or fibrated samples. In-field samples processing is very difficult and time-consuming, especially during harvest. Thus, the development of a rapid and efficient measurement technique which can be performed directly on stalk samples is highly desirable.

In this thesis, a new quality measurement method for fresh sugarcane stalk samples was developed using a visible and shortwave near infrared spectroradiometer (VNIRS) with the wavelength ranging from 350 to 1075 nm. A light-proof measurement box was developed and used as an instrument platform to evaluate the capability of the VNIRS to measure quality parameters of sugarcane samples. The box was used to determine quality parameters using two newly proposed scanning methods: the skin scanning method (SSM) and the cross sectional scanning method (CSSM). These methods were applied on both whole stalk and internode samples. No preparation mechanism was required prior to the quality measurement on stalk samples.

The selection of chemometrics methods used to optimise the regression models between spectral data and sugar content were also investigated. Partial least square (PLS) regression analysis with full cross validation (leave-one-out) technique was chosen to establish regression models between the spectral data and quality parameters. To improve the accuracy of the regression models, the spectral data was first pre-processed using the multiplicative scatter correction (MSC) method. Principal component analysis (PCA) was then used to extract useful information from the spectral data, decrease the noise and determine the optimum number of latent variables (LVs). The pre-processing methods, PLS and PCA exercises were run using Unscrambler V 9.6 software. The RPD (ratio of prediction to deviation) value was also used to evaluate the performance of the models.

For whole stalk samples, it was found that the R2 for SSM and CSSM were 0.82 and 0.68, respectively. The calibration models for the fibrated, juice and whole stalk samples were developed using quality values obtained by standard industry procedures. For internode samples, the R2 for SSM and CSSM were 0.91 and 0.87, respectively. The calibration models for internode samples were developed using °Brix values obtained from a handheld refractometer. The RPD values of the prediction models for internode samples by both SSM and CSSM were 2, indicating that these newly proposed methods can be used for coarse quantitative prediction purposes.

The variation of sugar content (°Brix) along the length of the stalks and internode samples were also assessed. The understanding of these variations can provide a foundation toward the design and development of the quality measurement system in the field. In this study, sugar content was found to vary significantly between the first and last internodes, with their average °Brix values being 22.2 and 7.6, respectively. The variation of sugar content between node and internode areas was 7.6% (SSM method) and 8.7% (CSSM method), respectively.

To demonstrate the possible applications of the proposed methods on a harvester, a basic calculation and conceptual design for a proposed in-field quality measurement system was outlined using the VNIRS mounted on top of the elevator conveyor. The proposed system had the potential to sense billet samples based on SSM either by directly scanning the moving billets on the elevator or by scanning the billets supplied by a sampling mechanism using a vacuum system. This theoretical design has shown that it is technically possible to develop a quality measurement system on a sugarcane harvester. However, more work needs to be done before this proposed method can be successfully mounted on a harvester.

Overall, it is concluded that the accuracy of the new measurement methods based on stalk samples using portable and low-cost VNIRS developed in this thesis is adequate. The proposed methods have significant potential uses as a tool for measuring sugarcane quality parameters from stalk samples in the field.