A Multimodal hyperspectral dataset of cocoa beans with physicochemical annotation

This work aimed to explore the possibility of predicting total fat content in whole dried cocoa beans at a single bean level using hyperspectral imaging (HSI).170 beans randomly selected from 17 batches were individually analysed by HSI and by reference methodology for fat quantification. Both whole (i.e. in-shell) beans and shelled seeds (cotyledons) were analysed. Partial Least Square (PLS) regression models showed good performance for single shelled beans (R2 = 0.84, external prediction error of 2.4%). For both in-shell beans a slightly lower prediction error of 4.0% and R2 = 0.52 was achieved, but fat content estimation is still of interest given its wide range. Beans were manually segregated, demonstrating an increase by up to 6% in the fat content of sub-fractions.HSI was shown to be a valuable technique for rapid, non-contact prediction of fat content in cocoa beans even from scans of unshelled beans, enabling significant practical benefits to the food industry for quality control purposes and for obtaining a more consistent raw material.

Introduction: Due to the fact that the presence of high doses of aflatoxin in agricultural products such as cocoa beans is unacceptable in terms of national and international standards, appropriate quality control tests should prevent such products to entering in the process of processing cocoa beans. Conventional methods of detecting aflatoxins such as ELISA and HPLC are very time consuming, expensive and require expertise, so replacing these tests with non-destructive and rapid methods such as near-infrared spectroscopy can increase the detection efficiency. Brado et al. (Berardo et al., 2005) used infrared spectroscopy to evaluate and diagnose Fusarium verticillium, which produces fumonicin toxin in maize. Manvar et al. (Mohammadi Manvar, 2015) used transmission and reflection Infrared spectroscopy to detect aflatoxin levels in Iranian pistachios. Singh et al. (2012) used hyperspectral imaging in the range of 700-1100 nm to detect fungal contamination of Penicillium SPP, Aspergillus Glaucus, and Aspergillus Niger in wheat. Kandpal et al. (Kandpal et al., 2015) in a research work using hyperspectral imaging in the range of 700-1100nm classified grains of maize contaminated with aflatoxin toxin using PLS-DA into five groups. In current study, an attempt was made to detect the amount of aflatoxin in cocoa beans using infrared spectroscopy and to classify healthy and infected beans into groups.Materials and methods: In this research, 180 cocoa beans, each weighing 1 gram, were selected to do analyses. One mg of aflatoxin B1 powder (A. flavus, A 6636, Sigma-Aldrich, St. Louis, Mo USA) was prepared from Sigma Aldrich representative in Iran and by dissolving this powder in absolute ethanol and concentrations of 20µg/kg, 500µg/kg was obtained as mentioned. For cocoa bean spectroscopy, a near infrared spectrometer in Shiraz University Central Laboratory (NIRS XDS Rapid Content Analysis) was used, which has the ability to spectroscopy in the range of 400-2500 nm. PLS-DA method was used to classify aflatoxin-infected samples from healthy samples. All 180 experimental samples were divided into two groups of training (120 samples) and test (60 samples) and the constructed model was first calibrated with training values and then evaluated with test data. Due to the fact that some noise is always stored in the spectral data and in order to remove this noise, a series of mathematical pretreatment, including: first and second derivatives was used (Chen et al., 2013; Nicolai et al., 2007).Results & Discussion: Comparing the average amount of infrared reflection spectrum, it is revealed that healthy grains have less reflection intensity than infected grains. Also, there are a number of local maximums and minimums where the difference in reflective intensity is more pronounced than elsewhere, and this phenomenon is due to the different concentrations of toxins in cocoa beans. After applying the second Savitzie Golay derivative pretreatment and performing PLS-DA classification using two latent variables, the distinction between classes can be clearly seen. The separation rate of the samples on the second LV is more specific, however, the second and first class samples in this LV have a closer score to each other. The peaks observed at 1440 nm and 1482 nm according to the first Everton O-H bond can be related to fungal contamination (Berardo et al., 2005; C. D. Sirisomboon et al., 2013). Also, the peak at 1838 nm is related to the tensile C-H bond, which can be related to the CH2 groups. According to the results obtained from the calibration, cross-validation and testing sections, it is determined that the degree of calibration error (ER) and the degree of error-free calibration (NER) in the pretreatment mode with the second-order derivative of Savitz Golay are the lowest and highest values, respectively. Also, in this pretreatment for the calibration model and testing, the specificity index for the first, second and third class samples are equal to 1.00, which means that all classes are correctly classified. In the cross-validation model, the value of the specificity index for the third class (samples with 500 ppb contamination) is equal to 97%. This indicates that 97% of infected seeds are correctly classified in the third group and only 3% in the other groups are incorrectly classified.Conclusion: The present study demonstrates the feasibility of near-infrared spectroscopy to identify and classify cocoa beans contaminated with aflatoxin. The results showed that the coefficients of independent variables (spectral wavelengths including 1440, 1482 and 1838 nm) decreased according to increasing in the concentration of toxin. Finally, it can be said that the method of detecting aflatoxin contamination using infrared spectroscopy is an efficient, non-destructive and fast method.

The aim of the current work was to use hyperspectral imaging (HSI) in the spectral range 1000–2500 nm to quantitatively predict fermentation index (FI), total polyphenols (TP) and antioxidant activity (AA) of individual dry fermented cocoa beans scanned on a single seed basis, in a non-destructive manner. Seventeen cocoa bean batches were obtained and 10 cocoa beans were used from each batch. PLS regression models were built on 170 samples. The developed HSI predictive models were able to quantify three quality-related parameters with sufficient performance for screening purposes, with external validation R2 of 0.50 (RMSEP = 0.27, RPD = 1.40), 0.70 (RMSEP = 34.1 mg ferulic acid g−1, RPD = 1.77) and 0.74 (60.0 mmol Trolog kg−1, RPD = 1.91) for FI, TP and AA, respectively. The calibrations were subsequently applied at a single bean and pixel level, so that the distribution was visualised within and between single seeds (chemical images). HSI is thus suggested as a promising approach to estimate cocoa bean composition rapidly and non-destructively, thus offering a valid tool for food inspection and quality control.

Premium cocoa beans are important for fine chocolate manufacturing. However, cocoa beans may suffer from improper management practices that reduce their final quality. The objective of this study was to establish a non-invasive and high throughput grading system for cocoa bean (whole seeds) using hyperspectral imaging technique, in combination with advanced machine learning methods. Six hundred cocoa beans were collected and scanned using a HySpex Classic SWIR camera covering the spectral range from 970 to 2500 nm, with a spatial resolution of 250 μm, and a spectral sampling of 5.45 nm. Each bean was then graded using cut test methodology, the internationally recognized standard procedure in the market for cocoa trade. A maximum entropy multiclass classification model was built based on the hyperspectral cube and results from the cut test. Cocoa beans were identified into different classes: good beans, under-fermented beans, slaty beans, and other low-quality beans. For the most critical classes (good, under-fermented and slaty), a classification accuracy close to 80% was achieved without having to cut the beans open. The classification model can also distinguish other defects such as germination, over-fermentation, mold, and white beans. The proposed hyperspectral solution can significantly increase the onsite evaluation capabilities for large number of samples, potentially applicable to full batches of cocoa beans. The analysis of all beans in a batch can provide a more reliable assessment of the overall quality, compared to the results traditionally obtained from cut tests using small sample sets from batches of several tons.

Cocoa beans are the most important raw material for the chocolate industry and an essential product for the economy of tropical countries such as Colombia. Their price mainly depends on their quality, which is determined by various aspects, such as good agricultural practices, their harvest point, and level of fermentation. The entities that regulate the international marketing of cocoa beans have been encouraging the development of new classification methods that, compared to current techniques, could save time, reduce waste, and increase the number of evaluated beans. In particular, hyperspectral images are a novel tool for food quality control. However, studies that have examined some quality parameters of cocoa using spectroscopy also involve the chemical evaluation of cocoa powder and liquor and the interior of the beans, which implies an invasive analysis, longer times, and waste generation. Therefore, in this paper, we assess the quality of cocoa beans based on their level of fermentation using a noninvasive system to obtain hyperspectral information, as well as fast image processing and spectral classification techniques. We obtained hyperspectral images of 90 cocoa beans in the range between 350 and 950 nm in an optical laboratory. In addition, each cocoa bean was classified according to its fermentation level: slightly fermented (SF), correctly fermented (CF), and highly fermented (HF). We compared this classification with that carried out by experts from the Colombia National Federation of Cocoa Growers and reported in the Colombian technical standard No. 1252. The results show that the level of fermentation of dried cocoa beans can be estimated using noninvasive hyperspectral image acquisition and processing techniques.

A hyperspectral image (HSI) consists of band images that capture the response from numerous light wavelengths in the electromagnetic spectrum. Therefore, HSI has a higher spectral resolution than RGB and grayscale images, which gives HSI enhanced characteristics that are useful in solving various real-world problems. In particular, hyperspectral imaging has been used for analysis and interpretation of material-specific properties in an image. Traditionally, most of the research on this topic has focused on the remote sensing. With the development of sensor technology, HSI has been introduced to other areas including scene understanding, medicine, food quality assessment etc. With the wide adoption of HSI, there is a demand to more effectively understand and exploit the information in HSI by exploring both intrinsic and extrinsic properties of HSI. The intrinsic properties in HSI consist of inherent information in the image due to various optical phenomenons. In contrast, the extrinsic property consists of compound information driven from intrinsic information. For this thesis, we focus mainly on how different properties in a hyperspectral image can contribute towards solving computer vision problems effectively. Although extrinsic properties of HSI can be explored for various research areas, we focus out research towards 3D computer vision due to its broad potential in both theoretical research and real-world applications. In this thesis, three main technical approaches are presented. Our first proposed approach focuses on the 3D reconstruction problem. We exploited spectral details in HSI to get the 3D structure, and we call it as structure from spectra. This is a challenging problem because 3D models reconstructed from different spectral bands demonstrate different structural properties. Our proposed method first generates 3D point sets from images at each wavelength using the typical structure from motion approach. A structural descriptor is developed to characterize the spatial relationship between the points, which allows robust point matching between two 3D models at different wavelengths. Then a 3D registration method is introduced to combine all band-level models into a single and complete hyperspectral 3D model. As far as we know, this is the first attempt in reconstructing a complete 3D model from hyperspectral images. This work allows fine structural-spectral information of an object to be captured and integrated into the 3D model, which can be used to support further research and applications. Our second method focuses on relative depth estimation. Here we exploited chromatic aberration and defocus blur in the monocular HSI to get depth cues. We propose that change in focus across band images of HSI due to chromatic aberration and band-wise defocus blur can be integrated for depth estimation. Then, novel methods are developed to estimate sparse depth maps based on different integration models. After that, by adopting manifold learning, an effective objective function is designed to combine all sparse depth maps into a final optimized sparse depth map. Lastly, a new dense depth map generation approach is proposed, which extrapolates sparse depth cues by using material-based properties on graph Laplacian. Experimental results show that our methods successfully exploit HSI properties to generate depth cues. We also compare our method with state-of-the-art RGB image-based approaches, which shows that our methods produce better sparse and dense depth maps than those from the benchmark methods. Our third novel approach deals with the occluded object detection problem. We use depth cues and material based information (reflectance) to get extrinsic property of continuous labels for detecting objects in the occluded environment. We propose the first method that exploits the unique fine reflectance properties of HSI to explore joint spectral and spatial information in HSI for occlusion detection. Specifically, our approach combines material distribution with depth cues extracted from a monocular HSI. The material distribution is estimated using a blind unmixing method based on hierarchical clustering. Then depths of objects are calculated using defocus blur and chromatic aberration. Finally, we combine both information using graph Laplacian to detect occluded objects. Lastly, in the Appendix, two supplementary methods are presented, which deals with spectral-spatial feature detection and object boundary detection problems in HSI. These two methods were partially derived from the research related to this thesis. They employ reflectance with structural and geometric cue present in HSI in different ways to achieve the goal.

n/a.

Elemental mapping of cocoa beans with laser-induced breakdown spectroscopy.

Eleven important medicinal plants generally used by the people of Turkey for the treatment of common cold have been studied for their mineral contents. Eleven minor and major elements (essential, non-essential and toxic) were identified in the Asplenium adiantum-nigrum L. , Althaea officinalis L. , Verbascum phlomoides L., Euphorbia chamaesyce L., Zizyphus jujube Miller, Peganum harmala L., Arum dioscoridis Sm., Sambucus nigra L., Piperlongum L., Tussilago farfara L. and Elettaria cardamomum Maton by employing flame atomic absorption and emission spectrometry and electro-thermal atomic absorption spectrometry. Microwave digestion procedure for total concentration was applied under optimized conditions for dissolution of medicinal plants. Plant based biological certified reference materials (CRMs) served as standards for quantification. These elements are found to be present in varying concentrations in the studied plants. The baseline data presented in this work can be used in understanding the role of essential, non-essential and toxic elements in nutritive, preventive and therapeutic properties of medicinal plants.

17 - Improving the detection of cocoa bean fermentation-related changes using image fusion

Hyperspectral imaging technology has been broadly applied in remote sensing because it collects echoed signals from across the electromagnetic (EM) spectrum and provides fruitfully helpful information. However, the processing or transformation of high-data-volume hyperspectral images, also viewed as snapshots varying with the EM spectrum, burdens the hardware resources, especially for the high spectral resolution and spatial resolution cases. To tackle this challenge, a novel reduced-order method based on the dynamic mode decomposition (DMD) algorithm is presented here to analyze hyperspectral images. The method decomposes the spatial-spectral hyperspectral images in terms of spatial dynamic modes and corresponding spectral patterns. Then, these spatial-spectral patterns are utilized to recover the raw hyperspectral images. Our proposed approach is benchmarked by the actual hyperspectral images measured at the Salinas scene. It is demonstrated that the proposed approach can represent the hyperspectral images with a low-rank model in spectral dimension. Our proposed approach could provide a useful tool for the model order reduction of hyperspectral images.

Atomic absorption spectroscopy (AAS) is a technique which is employed for the determination of 70–80 elements both qualitatively and quantitatively. The methods for determination are categorized as absorption, emission, or fluorescence with the detection limit ranging to ppm levels. Atomic absorption spectroscopy (AAS) is a well-known and accepted technique for determining elements up to the trace (μg/ml) and ultra-trace (sub-μg/ml) levels with great accuracy and acceptable precision. In this chapter, basic principles of AAS and techniques for the processing of samples like dry ashing, wet digestion, and microwave digestion have been discussed. The design and working of various components of flame AAS have been covered. The performance of flame and graphite furnace AAS has been compared. Various practical considerations for analyzing a sample on AAS have been elaborated. Apart from this, various types of interferences like spectral and nonspectral interferences which occur in AAS have also been discussed. The principle of hydride generation AAS (HGAAS) and mercury cold vapor generation has also been explained. The basic principle, design, and working of inductively coupled plasma-mass spectrometry (ICP-MS) and inductively coupled plasma-atomic emission spectroscopy (ICP-AES) have also been covered. Flame photometry has been compared with AAS. Applications of AAS in medicine, biochemistry, toxicology, water and air analysis, and agricultural and food stuffs have also been covered. A standard protocol for the determination of iron content by atomic absorption spectrometric method in infant milk substitute and determination of sodium and potassium content in milk samples by flame photometry have also been provided.KeywordsFlame and graphite furnace AASInductively coupled plasma-mass spectrometry (ICP-MS)Inductively coupled plasma-atomic emission spectroscopy (ICP-AES)Hydride generation AAS (HGAAS)Mercury cold vapor generation

Cutaneous ulcer caused by Leishmaniasis is a neglected disease which is more common in low-income areas. The main challenge in this disease is its diagnosis; the lack of specialized physicians makes its diagnosis difficult since quite often it can be miss-diagnosed with other type of skin ulcers such us venous, diabetes, and others. Given the previous mentioned facts, cutaneous ulcers caused by cutaneous Leishmaniasis require for the creation of novel tools that could assist its diagnosis. Hyperspectral and multispectral images measure the radiance reflected and emitted by a surface in hundreds or tens of spectral bands along the electromagnetic spectrum. This type of systems has been used for the analysis of cutaneous pathologies such us cancer, vitiligo, melasma, among others. With a set of classified hyperspectral images of cutaneous ulcers caused by different pathologies, it is possible to create an algorithm based on a multilayer neural network in order to achieve a classification of different types of ulcers. In this article we present the design of a feed-forward artificial neural network for the classification of cutaneous ulcers’ hyperspectral images in 4 kind of causes: occlusive vasculopathy, venous, Leishmaniasis, and diabetic. As result, a neural network structure is obtained that achieves a percentage of success higher than 72% in the classification of data.

Determination of trace elements and calcium in bone of the human iliac crest by atomic absorption spectrometry

Retrieving the reflectance spectrum from objects is an essential task for many classification and detection problems, since many materials and processes have a unique spectral behavior. In many cases, it is highly desirable to capture hyperspectral images due to the high spectral flexibility. Often, it is even necessary to capture hyperspectral videos or at least to be able to record a hyperspectral image at once, also called snapshot hyperspectral imaging, to avoid spectral smearing. For this task, a high-resolution snapshot hyperspectral camera array using a hexagonal shape is introduced. The hexagonal array for hyperspectral imaging uses off-the-shelf hardware, which enables high flexibility regarding employed cameras, lenses, and filters. Hence, the spectral range can be easily varied by mounting a different set of filters. Moreover, the concept of using off-the-shelf hardware enables low prices in comparison to other approaches with highly specialized hardware. Since classical industrial cameras are used in this hyperspectral camera array, the spatial and temporal resolution is very high, while recording 37 hyperspectral channels in the range from 400 to 760nm in 10nm steps. As the cameras are at different spatial positions, a registration process is required for near-field imaging, which maps the peripheral camera views to the center view. It is shown that this combination using a hyperspectral camera array and the corresponding image registration pipeline is superior in comparison to other popular snapshot approaches. For this evaluation, a synthetic hyperspectral database is rendered. On the synthetic data, the novel approach, to our knowledge, outperforms its best competitor by more than 3dB in reconstruction quality. This synthetic data is also used to show the superiority of the hexagonal shape in comparison to an orthogonal-spaced one. Moreover, a real-world high-resolution hyperspectral video database with 10 scenes is provided for further research in other applications.