Portable Hyperspectral Imaging Research Articles

Fifteenth century Florentine mural investigated in situ with VNIR Hyperspectral Imaging and NIR Photography supports interpretation as a bloodletting scene

This study provides new data which suggest a novel interpretative hypothesis not only on the specific painting, but on the use of bloodletting as medical practice in the Florentine Quattrocento. As a part of a cycle of frescoes devoted to the Seven Corporal Works of Mercy, the examined lunette depicts the “Visit to the sick” in a domestic interior, but it has never been considered as an historical document of precise medical practices. The scene’s definitive interpretation is still unresolved because of the uncertainty of some iconographic details. A campaign of in-situ and non-invasive technical investigations was performed to retrieve possible traces of previous details today concealed. The technical solutions adopted to implement the measurements campaign are illustrated, as an experimental example for remote sensing inspection of mural paintings in-situ. The position of the painting high up on a wall of an historical venue led to opting for stand-alone optical imaging techniques which could operate in remote sensing mode. By combining the use of portable Hyperspectral Imaging with Near Infrared photography a set of detailed images could be obtained that highlighted details not otherwise detectable. Focused on the objects held by the persons present, the analysis of the mural of Visit of the Buonomini in her Lying in Bed, the gift of swaddling cloth could be a tourniquet, shadows of folds of a blanket a thumb lancet, and an object held a blood collection bowl, supported the hypothesis that it could be a medieval bloodletting scene.

Non-invasive characterization of color variation in plant-based meat burgers using portable hyperspectral imaging device and multivariate image analysis

The excessive color variation in plant-based meat burgers (PB) during storage, arising from the interplay of multiple ingredients, suggests the acute need for real-time color monitoring techniques. This study reports the use of a smart-portable hyperspectral imaging device in the visible-near-infrared coupled with multivariate data and image analysis for the prediction of color in PB samples of different formulations during storage. Principal Components Analysis (PCA) and Partial Least Squares Regression (PLSR) were applied to explore the spectral features and predict the CIE - L*, a*, b*. PCA on spectral data showed successful spatial separation of PB samples based on color stability. The PLSR models in the full spectral range possessed good prediction accuracies for L* (R2P = 0.88, RMSEP = 0.94), a* (R2P = 0.98, RMSEP = 0.15), b* (R2P = 0.92, RMSEP = 0.31). Prediction capabilities of simplified PLSR models were improved for a* (R2P = 0.99, RMSEP = 0.09) with B-iPLS, b* (R2P = 0.95, RMSEP = 0.23) with B-iPLS. Pixel-by-pixel prediction maps unveiled the pattern of color deterioration concerning individual base ingredients in the PB samples. The results unravel the feasibility of developing a real-time multi-spectral technique for quantification and visualization of color evolution in plant-based meat analogs.

Utilization of Hyperspectral Imaging with Chemometrics to Assess Beef Maturity.

There is a growing demand from consumers for more assurance in premium food products such as beef and especially steak. The quality of beef steak is primarily dictated by the maturation which ultimately influences its taste and flavor. These enhanced qualities have resulted in steak becoming a premium product that consumers are willing to pay a premium price for. A challenge, however, is analyzing the maturity of beef by traditional analytical techniques. Hyperspectral imaging (HSI) is a methodology that is gaining traction mainly due to miniaturization, improved optics, and software. In this study, HSI was applied to wet aged beef supplied at various stages of maturity, with spectral data generated using a portable hyperspectral camera. Two trials were conducted over a five-month period: (i) proof of principle and (ii) a bespoke sampling trial for the industry. With the support of industry participation, all samples were sourced from a highly reputable UK/Ireland supplier. To enhance data interpretation, the spectral data collected were combined with multivariate analysis. A range of chemometric models were generated using unsupervised and supervised methods to determine the maturity of the beef, and external validation was performed. The external validation showed good accuracy for "unknown samples" tested against the model set and ranged from 74 to 100% for the different stages of maturity (20, 30, and 40 days old). This study demonstrated that HSI can detect different maturity timepoints for beef samples, which could play an important role in solving some of the challenges that the industry faces with ensuring the authenticity of their products. This is the first time that portable HSI has been coupled with chemometric modeling for assessing the maturity of beef, and it can serve as a model for other food authenticity and quality applications.

Portable hyperspectral imaging coupled with multivariate analysis for real-time prediction of plant-based meat analogues quality

Growing demand for plant-based meat analogues and their quality changes during storage necessitates real-time quality monitoring and chemical mapping. This study reports the use of portable hyperspectral imaging in the visible-near-infrared (VNIR) range (400–1000 nm) coupled with multivariate analysis for the prediction of quality traits (color, moisture, pH, and hardness) in plant-based meat burgers (PBMB) of different formulations ranging from 10 % to 30 % textured vegetable protein (TVP) and 5 % to 10 % pea protein (PP) during a 14-day storage period and map the quality distribution on each pixel. PCA was applied to explore both spectral (PCA-S) and measured quality data (PCA-M), while PLSR was applied for prediction modeling. ANOVA and PCA-M results showed the instability of PBMB quality even over a short storage period. PCA-S showed successful spatial separation of PBMB samples based on storage days, TVP, and PP concentrations. The PLSR models possessed good prediction accuracies for redness (R2P = 0.92, RMSEP = 0.74), yellowness (R2P = 0.93, RMSEP = 0.44), moisture (R2P = 0.90, RMSEP = 0.81 % Wb), and hardness (R2P = 0.81, RMSEP = 0.87 N). Overall, the prediction results and distribution maps unravel the feasibility of VNIR-HSI coupled with chemometrics for real-time quality inspection in plant-based food production.

LeafSpec-Dicot: An Accurate and Portable Hyperspectral Imaging Device for Dicot Leaves

Soybean is one of the world's most consumed crops. As the human population continuously increases, new phenotyping technology is needed to develop new soybean varieties with high-yield, stress-tolerant, and disease-tolerant traits. Hyperspectral imaging (HSI) is one of the most used technologies for phenotyping. The current HSI techniques with indoor imaging towers and unmanned aerial vehicles (UAVs) suffer from multiple major noise sources, such as changes in ambient lighting conditions, leaf slopes, and environmental conditions. To reduce the noise, a portable single-leaf high-resolution HSI imager named LeafSpec was developed. However, the original design does not work efficiently for the size and shape of dicot leaves, such as soybean leaves. In addition, there is a potential to make the dicot leaf scanning much faster and easier by automating the manual scan effort in the original design. Therefore, a renovated design of a LeafSpec with increased efficiency and imaging quality for dicot leaves is presented in this paper. The new design collects an image of a dicot leaf within 20 s. The data quality of this new device is validated by detecting the effect of nitrogen treatment on soybean plants. The improved spatial resolution allows users to utilize the Normalized Difference Vegetative Index (NDVI) spatial distribution heatmap of the entire leaf to predict the nitrogen content of a soybean plant. This preliminary NDVI distribution analysis result shows a strong correlation (R2 = 0.871) between the image collected by the device and the nitrogen content measured by a commercial laboratory. Therefore, it is concluded that the new LeafSpec-Dicot device can provide high-quality hyperspectral leaf images with high spatial resolution, high spectral resolution, and increased throughput for more accurate phenotyping. This enables phenotyping researchers to develop novel HSI image processing algorithms to utilize both spatial and spectral information to reveal more signals in soybean leaf images.

Geographical origin differentiation and quality determination of saffron using a portable Hyperspectral imaging system

A portable Hyperspectral Imaging (HSI) system (400–1000 nm) coupled with chemometrics techniques was developed to identify the geographical origin and to determine the Crocins content of saffron rapidly and non-destructively. Reflectance spectra of 160 saffron samples from two geographical origins were captured and preprocessed using the Savitzky–Golay (SG) algorithm. PCA as an unsupervised clustering and PLS-DA, RBF, and MLP as supervised techniques were applied for the classification and geographical origin differentiation of the samples. Afterward, the sample’s Crocins content was measured and their associations with their spectra were modeled using PLSR, RBF, and MLP models. The PCA results indicated that saffron samples were originally distributed in two different clusters. PLS-DA model with a spectral range from 400 to 1000 nm achieved the best clustering performance (Sensitivity and Specificity: 100 %). The RBF model showed good predictive capabilities in the saffron Crocins content prediction and quality evaluation (R2cal > 0.94 and R2pred > 0.83).

Global calibration for non-targeted fraud detection in quinoa flour using portable hyperspectral imaging and chemometrics

Quinoa is one of the highest nutritious grains, and global consumption of quinoa flour has increased as people pay more attention to health. Due to its high value, quinoa flour is susceptible to adulteration. Cross-contamination between quinoa flour and other flour can be easily neglected due to their highly similar appearance. Therefore, detecting adulteration in quinoa flour is important to consumers, industries, and regulatory agencies. In this study, portable hyperspectral imaging in the visible near-infrared (VNIR) spectral range (400–1000 nm) was applied as a rapid tool to detect adulteration in quinoa flour. Quinoa flour was adulterated with wheat, rice, soybean, and corn in the range of 0–98% with 2% increments. Partial least squares regression (PLSR) models were developed, and the best model for detecting the % authentic flour (quinoa) was obtained by the raw spectral data with R2p of 0.99, RMSEP of 3.08%, RPD of 8.77, and RER of 25.32. The model was improved, by selecting only 13 wavelengths using bootstrapping soft shrinkage (BOSS), to R2p of 0.99, RMSEP of 2.93%, RPD of 9.18, and RER of 26.60. A visualization map was also generated to predict the level of quinoa in the adulterated samples. The results of this study demonstrate the ability of VNIR hyperspectral imaging for adulteration detection in quinoa flour as an alternative to the complicated traditional method.

Hyperspectral imaging for assessment of total fat in salmon fillets: A comparison between benchtop and snapshot systems

Portable hyperspectral Imaging (HSI) based on snapshot sensors is ideal for on-line applications, with very fast data acquisition, small camera and flexibility for implementation. The objective of the study was to assess a snapshot HSI system for on-line assessment of salmon by bench marking it against a conventional hyperspectral system. Two different HSI systems: (a) linescan (550–1700 nm) and (b) snapshot (Vis: 470–630nm and Vis-NIR: 670–950 nm) were investigated. For each sample, one image was collected using linescan while two images per camera were collected using each snapshot followed by chemical analysis of samples to measure total fat content. Partial Least Squares Regression (PLSR) was successfully utilized to develop fat prediction models. Two strategies were followed: (1) Models for each dataset including stitched data from two snapshots and (2) HSI technology vs spectral range: new datasets were generated in such a way that both linescan and the device being compared (e.g., linescan vs Vis snapshot) had the same number of wavelengths in the similar wavelength range. For strategy (1) all models showed good prediction accuracy, yielding R p 2 in the range of 0.78–0.94 and SEP in the range of 1.58–3.10 except for Vis snapshot. For strategy (2) all models performed well, yielding R p 2 in the range of 0.77–0.94 and SEP in the range of 1.74–3.19 except for linescan vs Vis snapshot. Additionally, the stitched data model was used to generate fat prediction map of a salmon fillet and compared to its reference fat map. Results showed that snapshot HSI can be successfully used to predict fat in salmon fillets and have great potential for on-line applications. • • Hyperspectral linescan and snapshot cameras were compared. • • Both cameras successfully predicted total fat in salmon fillets via PLSR. • • Relevance of wavelength range was shown in predicting fat. • • PLSR model was used to generate fat prediction map of salmon fillets. • • Supporting evidence for developing hyperspectral system for industry is reported.

Assessment of microcirculatory changes in local tissue oxygenation after revascularization for peripheral arterial disease with the Hyperview®, a portable hyperspectral imaging device.

The Hyperview® is a hyperspectral camera, which can be used to assess the microcirculation of patients with peripheral arterial disease (PAD) and/or diabetes mellitus (DM). It measures local tissue oxygenation in concentrations of oxyhemoblobin (OXY), deoxyhemoglobin (DEOXY), and O2-saturation (O2-SAT) in arbitrary units. The aim of this validation study is to assess whether the Hyperview® is able to monitor microcirculatory changes after revascularization in patients with PAD. In this prospective observational cohort study, 50 patients with PAD were included who were scheduled for endovascular, hybrid, or open revascularization. The ankle-brachial index (ABI), systolic toe pressure (TP) (in case of DM), and a set of Hyperview® measurements of the plantar region were recorded before and after treatment. Changes in pre- and postoperative measurements were assessed with the paired t-test. Some 38 patients underwent endovascular revascularization and eight patients underwent hybrid or open vascular surgical revascularization. After revascularization, the ABI improved from 0.58 to 0.80 (p < 0.001). OXY increased from 72.6 to 77.8 (p = 0.134). DEOXY decreased from 69.1 to 55.0 (p < 0.001). O2-SAT increased from 51.3 to 58.1 (p < 0.001). TP improved from 50.6 to 61.9mmHg (p = 0.065) but was measured in only 16 patients. The Hyperview® is able to observe changes in the microcirculation after revascularization in terms of DEOXY and O2-SAT. The results of this study are a promising step into the validation of the Hyperview®.

Practical snapshot hyperspectral imaging with DOE

As the pursuit of snapshot spectral imaging continued to grow, traditional hyperspectral imaging systems have been too enormous and too slow to implement in real scenarios. Considering the portability and the demand for snapshots, in this study, we proposed a practical hyperspectral camera with a designed diffractive optical element (DOE). The designed DOE distinguished the incident spectrums and converged them into different point spread functions (PSFs) on the imaging plane. Utilizing the spectrally-varying PSF information, we engaged the iterative algorithm with the deep-learning model to reconstruct hyperspectral data. Experimental results demonstrated that the proposed system performed at least as well as the current methods and could achieve great spatial resolution and spectral accuracy in spectral imaging. This proposed system had good potential in portable hyperspectral imaging system.

Image Correction and In Situ Spectral Calibration for Low-Cost, Smartphone Hyperspectral Imaging

Developments in the portability of low-cost hyperspectral imaging instruments translate to significant benefits to agricultural industries and environmental monitoring applications. These advances can be further explicated by removing the need for complex post-processing and calibration. We propose a method for substantially increasing the utility of portable hyperspectral imaging. Vertical and horizontal spatial distortions introduced into images by ‘operator shake’ are corrected by an in-scene reference card with two spatial references. In situ light-source-independent spectral calibration is performed. This is achieved by a comparison of the ground-truth spectral reflectance of an in-scene red–green–blue target to the uncalibrated output of the hyperspectral data. Finally, bias introduced into the hyperspectral images due to the non-flat spectral output of the illumination is removed. This allows for low-skilled operation of a truly handheld, low-cost hyperspectral imager for agriculture, environmental monitoring, or other visible hyperspectral imaging applications.

Assessment of Strawberry Ripeness Using Hyperspectral Imaging

Portable hyperspectral imaging was used for field and indoor spectra acquisition of the strawberries at three ripeness stages: ripe, mid-ripe and unripe. The mean spectra were pre-processed by multiplicative scatter correction (MSC). Principal component analysis (PCA) was employed to generate score scatter plots and visualize score images for differentiating specific grouping of samples. Three methods, including X-loading weight, competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA), were applied to extract the effective wavelengths. Two classifiers, partial least squares – discriminant analysis (PLS-DA) and least squares – support vector machine (LS-SVM) were used for ripeness assessment. The results showed that the overall accuracy of all classifiers for field assessment ranged from 91.7% to 96.7%, slightly lower than for indoor assessment. Furthermore, the LS-SVM model combined with effective wavelengths with the CARS method performed better for field assessment of strawberry ripeness, providing an accuracy of 96.7%. It can be concluded that hyperspectral imaging can be used for real-time assessment of strawberry ripeness in the field.

Snapshot fiber spectral imaging using speckle correlations and compressive sensing.

Snapshot spectral imaging is rapidly gaining interest for remote sensing applications. Acquiring spatial and spectral data within one image promotes fast measurement times, and reduces the need for stabilized scanning imaging systems. Many current snapshot technologies, which rely on gratings or prisms to characterize wavelength information, are difficult to reduce in size for portable hyperspectral imaging. Here, we show that a multicore multimode fiber can be used as a compact spectral imager with sub-nanometer resolution, by encoding spectral information within a monochrome CMOS camera. We characterize wavelength-dependent speckle patterns for up to 3000 fiber cores over a broad wavelength range. A clustering algorithm is employed in combination with l1-minimization to limit data collection at the acquisition stage for the reconstruction of spectral images that are sparse in the wavelength domain. We also show that in the non-compressive regime these techniques are able to accurately reconstruct broadband information.

Portable Spectroscopy.

Until very recently, handheld spectrometers were the domain of major analytical and security instrument companies, with turnkey analyzers using spectroscopic techniques from X-ray fluorescence (XRF) for elemental analysis (metals), to Raman, mid-infrared, and near-infrared (NIR) for molecular analysis (mostly organics). However, the past few years have seen rapid changes in this landscape with the introduction of handheld laser-induced breakdown spectroscopy (LIBS), smartphone spectroscopy focusing on medical diagnostics for low-resource areas, commercial engines that a variety of companies can build up into products, hyphenated or dual technology instruments, low-cost visible-shortwave NIR instruments selling directly to the public, and, most recently, portable hyperspectral imaging instruments. Successful handheld instruments are designed to give answers to non-scientist operators; therefore, their developers have put extensive resources into reliable identification algorithms, spectroscopic libraries or databases, and qualitative and quantitative calibrations. As spectroscopic instruments become smaller and lower cost, "engines" have emerged, leading to the possibility of being incorporated in consumer devices and smart appliances, part of the Internet of Things (IOT). This review outlines the technologies used in portable spectroscopy, discusses their applications, both qualitative and quantitative, and how instrument developers and vendors have approached giving actionable answers to non-scientists. It outlines concerns on crowdsourced data, especially for heterogeneous samples, and finally looks towards the future in areas like IOT, emerging technologies for instruments, and portable hyphenated and hyperspectral instruments.

Design and Test of Portable Hyperspectral Imaging Spectrometer

We design and implement a portable hyperspectral imaging spectrometer, which has high spectral resolution, high spatial resolution, small volume, and low weight. The flight test has been conducted, and the hyperspectral images are acquired successfully. To achieve high performance, small volume, and regular appearance, an improved Dyson structure is designed and used in the hyperspectral imaging spectrometer. The hyperspectral imaging spectrometer is suitable for the small platform such as CubeSat and UAV (unmanned aerial vehicle), and it is also convenient to use for hyperspectral imaging acquiring in the laboratory and the field.

Use of a portable hyperspectral imaging system for monitoring the efficacy of sanitation procedures in produce processing plants

Cleaning and sanitation of production surfaces and equipment plays a critical role in lowering the risk of food borne illness associated with consumption of fresh-cut produce. Visual observation and sampling methods including ATP tests and cell culturing are commonly used to monitor the effectiveness of the cleaning procedures. This study tested the ability of a hand-held VIS hyperspectral imaging system to augment current monitoring methods. Multiple visits were made to two commercial fresh-cut processing facilities. Fluorescence-based detection and automated cycling among three wavelengths, 475, 520, and 675nm, proved best for detecting a range of anomalies. Numerous deficiencies in existing cleaning protocols were identified. Plant personnel were able to devise changes in the protocols that eliminated most of the detected problems without increasing the net expenditure for cleaning efforts. In addition, imaging identified sites where ATP test results were higher when compared to results for adjacent areas.

Development of a portable hyperspectral imaging system for monitoring the efficacy of sanitation procedures in food processing facilities

Cleaning and sanitation in food processing facilities is a critical step in reducing the risk of transfer of pathogenic organisms to food consumed by the public. Current methods to check the effectiveness of sanitation procedures rely on visual observation and sub-sampling tests such as ATP bioluminescence assays and culturing methods. To augment existing verification methods, a hand-held visible hyperspectral imaging device was developed. The device is capable of acquiring reflectance images using ambient lighting, and fluorescence responses to supplemental violet (405nm) excitation. To enhance the ability of detecting relatively low intensity fluorescence responses in the presence of ambient lighting, the device includes the ability to identify wavebands where the intensity of ambient lighting is relatively low. Valleys in ambient illumination intensity when using fluorescent lighting were found at around 475, 520, 570, and 675nm. A principal goal is to acquire data to support development of a commercially-viable, hand-held, imaging system.

Assessment of bacterial biofilm on stainless steel by hyperspectral fluorescence imaging

In this study, hyperspectral fluorescence imaging techniques were investigated for detection of microbial biofilm on stainless steel plates typically used to manufacture food processing equipment. Stainless steel coupons were immersed in bacterium cultures of nonpathogenic E. coli, Pseudomonas pertucinogena, Erwinia chrysanthemi, and Listeria innocua. Following a 1-week growth in rich medium tryptic soy broth (TSB) and M9 minimal medium with casamino acids (M9C), biofilm formations were evaluated using a recently developed portable hyperspectral fluorescence imaging system. Hyperspectral fluorescence images of the biofilm samples, in response to ultraviolet-A (320 to 400 nm) excitation, were acquired from 416 to 700 nm. Fluorescence images in the blue emission peak region exhibited the most contrast between biofilms and stainless steel coupons. On the basis of correlation analyses, two-band ratios compared with the single-band images enhanced the contrast between the biofilm forming area and untreated coupon surfaces. A two-band fluorescence ratio image, 444/588 nm, resulted in the greatest contrast between the biofilm formations and stainless steel coupon for the biofilms grown in M9C medium. TSB medium showed relatively high auto-fluorescence, and thus further investigation is needed to mitigate the contribution of strong TSB auto-fluorescence in detection of biofilms.

Citrus canker detection using hyperspectral reflectance imaging and PCA-based image classification method

Citrus canker is one of the most devastating diseases that threaten marketability of citrus crops. Technologies that can efficiently identify citrus canker would assure fruit quality and safety and enhance the competitiveness and profitability of the citrus industry. This research was aimed to investigate the potential of using hyperspectral imaging technique for detecting canker lesions on citrus fruit. A portable hyperspectral imaging system consisting of an automatic sample handling unit, a light source, and a hyperspectral imaging unit was developed for citrus canker detection. The imaging system was used to acquire reflectance images from citrus samples in the wavelength range between 400 and 900 nm. Ruby Red grapefruits with normal and various diseased skin conditions including canker, copper burn, greasy spot, wind scar, cake melanose, and specular melanose were tested. Hyperspectral reflectance images were analyzed using principal component analysis (PCA) to compress the 3-D hyperspectral image data and extract useful image features that could be used to discriminate cankerous samples from normal and other diseased samples. Image processing and classification algorithms were developed based upon the transformed images of PCA. The overall accuracy for canker detection was 92.7%. Four optimal wavelengths (553, 677, 718, and 858 nm) were identified in visible and short-wavelength near-infrared region that could be adopted by a future multispectral imaging solution for detecting citrus canker on a sorting machine. This research demonstrated that hyperspectral imaging technique could be used for discriminating citrus canker from other confounding diseases.

Plant classification for weed detection using hyperspectral imaging with wavelet analysis

The goal of this study is to develop a new weed detection method that can be applied for automatic mechanical weed control. For successful weed detection, plants must be classified into crops and weeds according to their species. In this study, we employed a portable hyperspectral imaging system. The hyperspectral camera can capture landscape images that include crops, weeds, and the soil surface, and can provide more extensive information than conventional red, green, and blue (RGB) images. Although RGB images consist of red, green, and blue wavebands, the obtained hyperspectral images consist of 240 wavebands of spectral information. Hyperspectral imaging is expected to provide powerful technology for agricultural sensing. In the initial step of this study, the image pixels of the plants (crop or weeds) were segmented from the background soil surface using Euclidean distance as the discriminant function. In the next step, the image pixels of the crop (sugarbeet) and weeds (four species) were classified using the difference in the spectral characteristics of the plant species. In this process, classification variables were generated using wavelet transformation for data compression, noise reduction, and feature extraction, and then stepwise linear discriminant analysis was applied. The validation results indicate that the developed classification method has potential for practical use.