Spatially resolved spectroscopy for detecting firmness and internal disorders of horticultural products

Abstract

Nondestructive detection of firmness and internal disorders are critical for New Zealand’s horticultural industry. Soft and internally defective products can result both in loss of revenue and the supplier’s reputation for quality. This thesis presents evaluation and development of spatially resolved spectroscopy (SRS) techniques aiming to improve the detection performance of these two quality parameters. Original contributions to literature include: (1) an assessment of an SRS reflectance system and a direct comparison of SRS and near-infrared spectroscopy (NIRS) on firmness prediction, (2) quantified understanding of NIRS on detecting internal disorders in terms of light transport and detection limitations, (3) accurate estimations of optical properties of healthy and rotten onion flesh, and (4) development of an SRS transmittance system for detecting onion rot, with the aid of finite element method (FEM) simulations. By applying SRS measurements on the same 100 ‘Royal Gala’ apples, this work showed that the NIRS interactance system slightly outperformed the SRS-based multispectral imaging (MSI) system when applied to the same samples. The cross validation prediction error was RMSECV = 6.99 N when compared with penetrometer firmness. Before the SRS system can be considered for commercial implementation, SRS firmness system must be further developed to accommodate the conditions and speed (10 samples/second) of a produce grader and demonstrate better performance than the existing online sensors. Internal disorders that are small and spatially confined inside produce are difficult to detect using NIRS due to its lack of spatial resolution, but improvement might be made with better locations of light illumination and detection on the sample, i.e. optical geometry. Two optical geometries were investigated for detecting small and localised vascular browning (VAB) disorders in ‘Braeburn’ apples. Monte Carlo (MC) simulations indicated that the 180° transmission geometry allowed the light to travel much further than that of the 90ºgeometry inside a sample. However, the sensor is still only sensitive to a small portion of the internal volume, which was further validated by the poor results from the actual measurements. The 180° geometry system could detect 80% of defective apples with a misclassification rate of 21% for healthy ones.iv An SRS system was then developed for detecting onion neck rots that are often small and localised to the stem end. The FEM software NIRFast was used to simulate light transport with absorption coefficients determined from juice transmittance measurements. The transmittance ratio of 728/805 nm was sensitive to the presence of the rot and was independent of the onion size. This ratio at different source locations formed a spatial profile, which could indicate the size and location of the rots. Finally, the optical system was implemented using a lock-in amplifier for simultaneous measurements of the light of two lasers at 728 nm and 805 nm. A mechanical scanning system was used to investigate multiple source locations. A preliminary test of the system showed a substantial improvement on typical NIRS performance levels. Future work aims to achieve a similar measurement scheme, while the sample is moving at high speed on a grader.