Rosalba Calvini

Data dimensionality reduction and data fusion for fast characterization of green coffee samples using hyperspectral sensors.

AbstractHyperspectral sensors represent a powerful tool for chemical mapping of solid-state samples, since they provide spectral information localized in the image domain in very short times and without the need of sample pretreatment. However, due to the large data size of each hyperspectral image, data dimensionality reduction (DR) is necessary in order to develop hyperspectral sensors for real-time monitoring of large sets of samples with different characteristics. In particular, in this work, we focused on DR methods to convert the three-dimensional data array corresponding to each hyperspectral image into a one-dimensional signal (1D-DR), which retains spectral and/or spatial information. In this way, large datasets of hyperspectral images can be converted into matrices of signals, which in turn can be easily processed using suitable multivariate statistical methods. Obviously, different 1D-DR methods highlight different aspects of the hyperspectral image dataset. Therefore, in order to investigate their advantages and disadvantages, in this work, we compared three different 1D-DR methods: average spectrum (AS), single space hyperspectrogram (SSH) and common space hyperspectrogram (CSH). In particular, we have considered 370 NIR-hyperspectral images of a set of green coffee samples, and the three 1D-DR methods were tested for their effectiveness in sensor fault detection, data structure exploration and sample classification according to coffee variety and to coffee processing method. Principal component analysis and partial least squares-discriminant analysis were used to compare the three separate DR methods. Furthermore, low-level and mid-level data fusion was also employed to test the advantages of using AS, SSH and CSH altogether. Graphical AbstractKey steps in hyperspectral data dimenionality reduction

Prediction of parameters related to grape ripening by multivariate calibration of voltammetric signals acquired by an electronic tongue

An electronic tongue (ET) consisting of two voltammetric sensors, namely a poly-ethylendioxythiophene modified Pt electrode and a sonogel carbon electrode, has been developed aiming at monitoring grape ripening. To test the effectiveness of device and measurement procedures developed, samples of three varieties of grapes have been collected from veraison to harvest of the mature grape bunches. The derived musts have been then submitted to electrochemical investigation using Differential Pulse Voltammetry technique. At the same time, quantitative determination of specific analytical parameters for the evaluation of technological and phenolic maturity of each sample has been performed by means of conventional analytical techniques. After a preliminary inspection by principal component analysis, calibration models were calculated both by partial least squares (PLS) on the whole signals and by the interval partial least squares (iPLS) variable selection algorithm, in order to estimate physico-chemical parameters. Calibration models have been obtained both considering separately the signals of each sensor of the ET, and by proper fusion of the voltammetric data selected from the two sensors by iPLS. The latter procedure allowed us to check the possible complementarity of the information brought by the different electrodes. Good predictive models have been obtained for estimation of pH, total acidity, sugar content, and anthocyanins content. The application of the ET for fast evaluation of grape ripening and of most suitable harvesting time is proposed.

Transferring results from NIR-hyperspectral to NIR-multispectral imaging systems: A filter-based simulation applied to the classification of Arabica and Robusta green coffee

Due to the differences in terms of both price and quality, the availability of effective instrumentation to discriminate between Arabica and Robusta coffee is extremely important. To this aim, the use of multispectral imaging systems could provide reliable and accurate real-time monitoring at relatively low costs. However, in practice the implementation of multispectral imaging systems is not straightforward: the present work investigates this issue, starting from the outcome of variable selection performed using a hyperspectral system. Multispectral data were simulated considering four commercially available filters matching the selected spectral regions, and used to calculate multivariate classification models with Partial Least Squares-Discriminant Analysis (PLS-DA) and sparse PLS-DA. Proper strategies for the definition of the training set and the selection of the most effective combinations of spectral channels led to satisfactory classification performances (100% classification efficiency in prediction of the test set).

Electronic eye for the prediction of parameters related to grape ripening

An electronic eye (EE) for fast and easy evaluation of grape phenolic ripening has been developed. For this purpose, berries of different grape varieties were collected at different harvest times from veraison to maturity, then an amount of the derived must was deposited on a white sheet of absorbent paper to obtain a sort of paper chromatography. Thus, RGB images of the must spots were collected using a flatbed scanner and converted into one-dimensional signals, named colourgrams, which codify the colour properties of the images. The dataset of colourgrams was used to build calibration models to relate the colour of the images with the phenolic composition of the samples - determined by reference analytical methods - and therefore to follow the ripening trend. Satisfactory calibration models were obtained for the prediction of the most important parameters related to phenolic ripening of grapes, such as colour index, tonality, total anthocyanins content, malvidin-3-O-glucoside and petunidin-3-O-glucoside.

Sparse-Based Modeling of Hyperspectral Data

Abstract One of the main issues of hyperspectral imaging data is to unravel the relevant, yet overlapped, huge amount of information contained in the spatial and spectral dimensions. When dealing with the application of multivariate models in such high-dimensional data, sparsity can improve the interpretability and the performance of the model. In this chapter, we will introduce the exploration of hyperspectral images using a sparse version of the well-known principal component analysis method to demonstrate how the derived models can reveal very useful spectral zones. In particular, we will present two practical applications related to different issues: the separation among groups of homogeneous samples and the identification of outlier pixels in the spatial domain. For both case studies, guidance to the identification of the proper level of sparsity will be provided and, furthermore, we will show how sparsity can improve the chemical interpretation of the results.