This PDF file contains the front matter associated with SPIE Proceedings Volume 12338, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

Radiometric calibration of all optical sensors is a key component to achieve high accuracy data, which is traceable to international standards and repeatable through time. This study introduces a Hyperspectral Imaging (HSI) system that RAL Space has developed and presents a thorough analysis of the radiometric reflectance calibration chain required for both static and aerial applications, where the HSI system is mounted on an Unoccupied Aerial System (UAS). RAL Space HSI system comprises two commercially available Ximea snapshot mosaic imagers, one sensitive in the visible (VIS) and the other in the near infrared (NIR) spectral ranges, that are coupled with a dedicated PC and customized LabVIEW code for the capture, process and storage of the raw hyperspectral information (hypercubes). Both imagers are optically characterized for their electrical gain and offset, the vignetting and stray light effects, while the spectral response of each band has also been determined. The corrected digital numbers (DN) are then translated into reflectance values with the use of reflectance standards. For the aerial surveys, an additional step has been developed into the calibration chain that corrects for any changes in ambient illumination during flight with the use of ground based upward-looking measured spectra. Future work will include any correction needed due to temperature dependencies of the calibration chain steps.

We present a novel hyperspectral imaging system working in the visible and in the short-wave infrared (SWIR) spectral region based on a Fourier-transform approach. The technology presents an exceptional light throughput, a high spatial resolution, a software adjustable spectral resolution, and a wide versatility of use. In this work, we illustrate a broad portfolio of applications both in the visible and in the SWIR regions, with particular focus on microscopy and biology, cultural heritage, and quality control for the agri-food sector, in collaboration with a vertical farm.

There is a growing interest in SWIR to solve many problems across a range of applications. Major advances have been made in Hyperspectral Imaging over the past few years. The data collected in a HIS system for analysis relies on a good camera, offering a combination of resolution, speed and sensitivity. In this presentation Mark Donaghy, from Raptor will discuss the many parameters in selecting the right camera and will then go through a series of example applications that customers are working on using Raptor SWIR cameras. He will discuss where SWIR technology / HIS goes next.

This paper presents the latest advances on Imec snapshot multispectral imagers based on either 3x3, 4x4 and 5x5 mosaic filter patterning on industry ready VIS/NIR and SWIR detectors. The mosaic patterns are implemented by means of high-transmission Fabry-Pérot interferometers processed using thin-film technology. Our snapshot multispectral imagers offer a spatial resolution of 640x480 pixels (SWIR) and 2048x1088 (VIS/NIR) down sampled according to the mosaic pattern to acquire data in nine (3x3), 16 (4x4) of 25 (5x5) spectral bands respectively. To achieve imaging at the native spatial resolution of the sensor, super resolution methods are available post-acquisition. Moreover, our compact USB-3 cameras of 260 gr (SWIR) and 27 gr (VNIR), without lens, reach an acquisition speed of up to 120 multispectral cubes/second and are therefore suitable for dynamic applications, high-speed inspection such as a conveyor belt or UAV inspection. The potential for snapshot cameras in a wide range of applications is showcased in this paper. We first show how applications on industrial quality inspection (chocolate gloss estimation) and precision agriculture (plant disease detection) achieve good discrimination potential in the VNIR range. Specifically, UAV inspection benefits from our compact camera size, low weight, and video capabilities. We then demonstrate the potential for plastic and textile recycling in the SWIR range and the benefit brought by both VNIR and SWIR ranges for people tracking under low visibility conditions. Finally, an application involving the joint use of a microscope and a multispectral camera system is presented for particle contamination exposure assessment. The suitable range is in this case application dependent.

HySpex is presenting an integrated solution for hyperspectral drill core imaging. The system’s mineral mapping capabilities are presented in close cooperation with renowned academic and industrial partners through the Center to Advance the Science of Exploration to Reclamation in Mining (CASERM) led by the Colorado School of Mines and Virginia Tech. Utilizing HySpex cameras covering the spectral range between 400 and 2500 nm, the system is capable of scanning full core boxes in seconds. Using Prediktera’s new Breeze-GEO Software, real-time mineral mapping of the highest quality is achieved. Apart from different interactive qualitative and quantitative data analysis tools offered by Breeze, the platform includes the publicly available USGS Material Identification and Classification Algorithm (MICA) for mineral identification, as well as the Minimum Wavelength Mapping (MWL) algorithm. The scanner’s capabilities are demonstrated using drill cores from the LaRonde-Penna deposit. The deposit is located within the Archean Abitibi greenstone belt of Ontario and Quebec, Canada, which is home to numerous Volcanogenic Massive Sulfide (VMS) deposits. LaRonde-Penna contains an endowment of 71 Mt of ore grading on average 3.9 g/t Au and economic grades of Zn, Cu and Pd. Because the deposits have been studied extensively over the past decades, cores from the deposit open up a unique opportunity for research and development.

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.

This work reviews and presents a comparison of hyperspectral imaging results when analyzing corrosion products in the ultraviolet (UV) range (250 nm to 500 nm), visible near-infrared (VNIR) range (400 to1000 nm) and shortwave-infrared range (900 to 2500 nm). In related and prior work in our group, corrosion products on steel have been detected using hyperspectral imaging in the VIS, NIR and SWIR regions of the spectrum. However, an extensive review of the academic literature has revealed that the hyperspectral response of corrosion in the UV has not been reported. To address this, we present our results of imaging corrosion products on metal substrates using our Headwall UV-VIS Hyperspectral imaging sensor. These results are contrasted with the same samples imaged using our Headwall VNIR E series and Headwall SWIR 640 Hyperspectral imaging sensors. Our initial results indicate that corrosion spectra in the UV are separable from those of steel, but that the VNIR is the most appropriate range for this type of determination.

We describe results from experiments investigating how hyperspectral data might be incorporated into autonomous inspections for offshore turbines, part of Dr SUIT– (Drone Swarm for Unmanned Inspection of Wind Turbines), a collaboration funded by InnovateUK (UKRI). Imagery and point measurements were captured of small turbine blades subjected to damage by abrasion, impact and UV exposure. The technique appears effective at classifying abrasion damage to a degree comparable with conventional inspection schemes. Impact damage could be classified as ‘lower’ or ‘higher’ energies. The blades designed resilience to UV meant that little change was detected in those tests.

The operation of hyperspectral imaging systems in industrial environments can be a challenge. In the nuclear industry, partially transparent elements such as gloveboxes or panels are often used to cover samples for protection against the risk of contamination. In practical terms, this means that the hyperspectral sensors can only capture data through partially transparent media, which interferes the vision between sensor and sample. Representative examples of these media are Polymethyl Methacrylate (PMMA) or acrylic and Polycarbonate (PC). In this work, we evaluate the effect that the transparent media can have on the data when captured under these conditions, where transparent materials are placed between sensor and sample. Experiments include hyperspectral images of the same samples captured with and without panel obstruction for a direct comparison of spectral responses, suggesting potential artificial intelligence techniques and methods to identify these effects and mitigate them.

In the brain tumour resection surgery, functional brain mapping (FBM) has been adapted in the contemporary neurosurgical workflow for improved surgical outcome. It allows neurosurgeons to identify and preserve brain regions with critical functions, thus to prevent post-surgical neurological deficits. However, there is no effective ways to directly visualize brain functions in real-time intraoperatively. A compact functional brain imaging device that can be more easily utilised during surgery is highly desired. We’ve been developing a multispectral imaging system (MSI) with the aim to fill this unmet clinical need. The device has been designed to detect brain functional activity by characterising light reflectance changes due to the blood oxygenation/flow fluctuation and neuronal membrane change. Co-localised MSI and fMRI measurements have been performed for detecting brain response to external electrical stimulations on preclinical animal models. Results suggested MSI could be able to identify the stimulation-evoked brain region as validated by the fMRI findings

Glioblastoma surgical resection is a problematic mission for neurosurgeons. Tumor complete resection improves patients healing chances and prognosis, whilst excessive resection could lead to neurological deficits. Nevertheless, surgeons' sight hardly traces the tumor's extent and boundaries. Indeed, most surgical processes result in subtotal resections. Histopathological testing might enable complete tumor elimination, though it is not feasible due to the time required for tissue investigation. Several studies reported tumor cells having unique molecular signatures and properties. Hyperspectral Imaging (HSI) is an emerging, non-contact, non-ionizing, label-free and minimally invasive optical imaging technique able to extract information concerning the observed tissue at the molecular level. Here, we exploited extensive data augmentation, transfer learning, the U-Net++ and the DeepLab-V3+ architectures to perform the automatic end-to-end segmentation of intraoperative glioblastoma hyperspectral images meeting competitive processing times and segmentation results concerning the gold-standard procedure. Based on ground truths provided by the HELICoiD framework, we dramatically improved HSIs processing times, enabling the end-to-end segmentation of glioblastomas targeting the real-time processing to be employed during open craniotomy in surgery, thus improving the gold-standard ML pipeline. We measured competitive inference times concerning the standard CUDA environment offered by MatLab 2020a. The HELICoiD fastest parallel version took 1.68 s to elaborate the most prominent image of the database, whilst our methodology performs segmentation inference in 0.29 ± 0.17 s, hence being real-time compliant concerning the 21 seconds constraint imposed on processing. Furthermore, we evaluated our segmentation results qualitatively and quantitatively regarding the ground truth produced by HELICoiD.

Hyperspectral Imaging (HSI) is a promising practice in research medicine due to its non-contact, non-invasive, non-ionizing, and label-free characteristics. Chromophores, such as haemoglobin and melanin, are responsible for the chemical structure of tissues and determine their spectral properties. Therefore, hyperspectral technologies might serve the role of tissue diagnosis, aiding physicians during surgical or clinical operations. Hence, hyperspectral cameras produce the data used by machine and deep learning algorithms to discriminate healthy from damaged tissues. Nevertheless, data quality remains an issue, especially concerning the small-sized medical dataset available to research. Here, we propose a hyperspectral imaging blueprint, designed to work with push broom sensors, representing one of the highest quality transducers to acquire spectral data. Indeed, push broom sensors only seize one scene line at a time, offering high spatial and spectral resolutions. It can work in any scenario, such as dermatological or surgical, involving a motionless subject. We designed the system to be affordable, open-source and robust. Therefore, it comprises Python libraries, an Arduino one board, a Nema17 stepper motor, its driver controller, and a recirculating ball screw for accurate movement. Furthermore, it offers a diode-based targeting system, attached to a 3D printed circular crown and built to hit the image capture and measure the right focusing distance. We equipped the blueprint with a graphical user interface to let physicians interact with the camera, accurately move it, and acquire the diagnostic data needed.

Hyperspectral cameras are capable of obtaining highly useful data for geology, agriculture, urban planning, and many other applications. Several satellite-based hyperspectral cameras are currently operational, providing hyperspectral data to various users. Even large instruments usually have relatively large ground sampling distance (GSD): 10m or larger in 400 to 1000nm range and 30m or larger in 900 to 2500nm range. GSD is even coarser in hyperspectral cameras for microsatellites. Based on the information from PRISMA 2021 Workshop and our customer’s feedback, the most requested feature for satellite-based hyperspectral cameras is significantly improved GSD. Also, there is a strong demand for smaller microsatellite-compatible hyperspectral cameras. Due to lower mission cost, such cameras can provide hyperspectral data to more users. Additionally, microsatellite constellations could provide swath and revisit time that would be impossible for a single large satellite. Creating a hyperspectral camera with acceptable Signal-to-Noise Ratio (SNR) and small GSD, that would be still compatible with a small platform, is a big challenge. Our approach has been to create a hyperspectral camera that would surpass the current limitations of small satellite platforms, and would provide data that, for some specifications, exceed what is available for free from large instruments. Our focus has been on providing significantly improved GSD, small spatial and spectral misregistration, while keeping acceptable spectral sampling and SNR. The instrument development has been funded by the Norwegian Space Agency. One of the proposed instruments has been selected by the Norwegian Space Agency as the primary payload on an upcoming Norwegian In-Orbit Demonstrator satellite.

Remote sensing applications are nowadays widely spread in various industrial fields, such as mineral and water exploration, geo-structural mapping, and natural hazards analysis. These applications require that the performance of image processing tasks, such as segmentation, object detection, and classification, to be of high accuracy. This can be achieved with relative ease if the given image has high spatial resolution as well as high spectral resolution. However, due to sensor limitations, spatial and spectral resolutions have an inherently inverse relationship and cannot be achieved simultaneously. Hyperspectral Images (HSI) have high spectral resolution, but suffer from low spatial resolution, which hinders utilizing them to their full potential. One of the most widely used approaches to enhance spatial resolution is Single Image Super Resolution (SISR) techniques. In the recent years, Deep Convolutional Neural Networks (DCNNs) have been widely used for HSI enhancement, as they have shown superiority over other traditional methods. Nonetheless, researches still aspire to enhance HSI quality further while overcoming common challenges, such as spectral distortions. Research has shown that properties of natural images can be easily captured using complex numbers. However, this has not been thoroughly investigated from the perspective of HSI SISR. In this paper, we propose a variation of a Complex Valued Neural Network (CVNN) architecture for HSI spatial enhancement. The benefits of approaching the problem from a frequency domain perspective will be answered and the proposed network will be compared to its real counterpart and other state-of-the-art approaches. The evaluation and comparison will be recorded qualitatively by visual comparison, and quantitatively using Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Spectral Angle Mapper (SAM).