Tea is the most consumed manufactured drink in the world. In recent years, various high end analytical techniques such as high-performance liquid chromatography have been used to analyse tea products. However, these techniques require complex sample preparation, are time consuming, expensive and require a skilled analyst to carry out the experiments. Therefore, to support rapid and non-destructive assessment of tea products, the use of near infrared (NIR) (950-1760 nm) hyperspectral imaging (HSI) for classification of six different commercial tea products (oolong, green, yellow, white, black and Pu-erh) is presented. To visualise the HSI data, linear (principal component analysis (PCA) and multidimensional scaling (MDS)) and non-linear (t-distributed stochastic neighbour embedding (t-SNE) and isometric mapping (ISOMAP)) data visualisation methods were compared. t-SNE provided separation of the six commercial tea products into three groups based on the extent of processing: minimally processed, oxidised and fermented. To perform the classification of different tea products, a multi-class error-correcting output code (ECOC) model containing support vector machine (SVM) binary learners was developed. The classification model was further used to predict classes for pixels in the HSI hypercube to obtain the classification maps. The SVM-ECOC model provided a classification accuracy of 97.41±0.16 % for the six commercial tea products. The methodology developed provides a means for rapid, non-destructive, in situ testing of tea products, which would be of considerable benefit for process monitoring, quality control, authenticity and adulteration detection.

Band selection plays an important role in hyperspectral data analysis as it can improve the performance of data analysis without losing information about the constitution of the underlying data. We propose a MIMR-DGSA algorithm for band selection by following the Maximum-Information-Minimum-Redundancy (MIMR) criterion that maximises the information carried by individual features of a subset and minimises redundant information between them. Subsets are generated with a modified Discrete Gravitational Search Algorithm (DGSA) where we definine a neighbourhood concept for feature subsets. A fast algorithm for pairwise mutual information calculation that incorporates variable bandwidths of hyperspectral bands called VarBWFastMI is also developed. Classification results on three hyperspectral remote sensing datasets show that the proposed MIMR-DGSA performs similar to the original MIMR with Clonal Selection Algorithm (CSA) but is computationally more efficient and easier to

In this study, ultra-violet (UV) and short-wave infra-red (SWIR) hyperspectral imaging (HSI) was used to measure the concentration of phenolic flavour compounds on malted barley that are responsible for smoky aroma of Scotch whisky. UV-HSI is a relatively unexplored technique that has the potential to detect specific absorptions of phenols. SWIR-HSI has proven to detect phenols in previous applications. Support Vector Machine Classification and Regression was applied to classify malts with ten different concentration levels of the compounds of interest, and to estimate the concentration respectively. Results reveal that UV-HSI is at its current development stage unsuitable for this task whereas SWIR-HSI is able to produce robust results with a classification accuracy of 99.8% and a squared correlation coefficient of 0.98 with a Root Mean Squared Error of 0.32 ppm for regression. The results indicate that with further testing and development, HSI may potentially be exploited in an industrial production environment.

Nondestructive phenolic compounds measurement and origin discrimination of peated barley malt using near-infrared hyperspectral imagery and machine learning. IEEE transactions on instrumentation and measurement [online], Early Access.

Biomedical engineering is a unique area that allows fusion between two distinct fields of engineering and medicine. The integration of efforts from both fields promises progress through acquisition of information from tissues, cells, and organs through non-invasive methods of assessment. Here we investigate the ability of a hyperspectral device in extracting data from tissues through the wavelength spectrum, in foreseeing its potential in clinical diagnostics by simplifying methods of examination by clinicians in detecting corneal injuries. Hyperspectral imaging using 400 to 1000nm visible wavelength was used to scan five porcine eyes (injured and non-injured). Images were saved in three dimensional images of rows, columns, and depth slices at 1200 to 1300x804x604 and were processed. All laboratory works were performed in accordance with the general risk assessment of University of Strathclyde. In our results, analysis of the images reveals significant cue between 500 to 800nm bands in differentiating between injured and noninjured parts of the eye.