Metasurface‐Encoded Single‐Pixel Hyperspectral Imaging

In order to quickly distinguish infertile eggs from fertile eggs, the hyperspectral imaging technology consisting of imaging and spectral information was used for detecting the fertile information of eggs. Before hatching eggs were incubated, a hyperspectral imaging system (wavelength between 400 to 1 000 nm) was used to acquire the images one-by-one manually. The characteristic information of ratios of length to short axis, elongation, roundness and the ratios of the yolk area to the whole area was extracted based on the images. The normalization method was used as the spectral data preprocessing, and then 155 spectral characteristic variables were extracted from 520 nm waveband through the correlation coefficient method. Principal component analysis (PCA) method was adopted to reduce the dimensions of image-spectrum fusion information; the top six principal components were extracted. Support vector machine (SVM) method was used to establish classification of fertile and infertile eggs models, which are based on image, spectrum and image-spectrum fusion information respectively. The accuracy rates of the SVM models were 84.00%, 90.00% and 93.00% respectively. The experimental results show that the model based on image-spectrum fusion information technology is superior to the single information model. Hyperspectral transmission imaging technology is effective and feasible to detect the fertile hatching eggs before incubation. Keywords: hyperspectral image, hatching eggs, information fusion, nondestructive detection DOI: 10.3965/j.ijabe.20150804.1672 Citation: Zhu Z H, Liu T, Xie D J, Wang Q H, Ma M H. Nondestructive detection of infertile hatching eggs based on spectral and imaging information. Int J Agric & Biol Eng, 2015; 8(4): 69－76.

Optical spectroscopy can be used to distinguish between healthy and diseased tissue. In this study, the design and testing of a single-pixel hyperspectral imaging (HSI) system that uses autofluorescence emission from collagen (400nm) and nicotinamide adenine dinucleotide phosphate (475nm) along with differences in the optical reflectance spectra to differentiate between healthy and thermally damaged tissue is discussed. The changes in protein autofluorescence and reflectance due to thermal damage are studied in ex vivo porcine tissue models. Thermal lesions were created in porcine skin (n=12) and liver (n=15) samples using an IR laser. The damaged regions were clearly visible in the hyperspectral images. Sizes of the thermally damaged regions as measured via HSI are compared to sizes of these regions as measured in white-light images and via physical measurement. Good agreement between the sizes measured in the hyperspectral images, white-light imaging, and physical measurements were found. The HSI system can differentiate between healthy and damaged tissue. Possible applications of this imaging system include determination of tumor margins during surgery/biopsy and cancer diagnosis and staging.

Hyperspectral imaging provides complex image data with spectral information from many fluorescent species contained within the sample such as the fluorescent labels and cellular or pigment autofluorescence. To maximize the utility of this spectral imaging technique it is necessary to couple hyperspectral imaging with sophisticated multivariate analysis methods to extract meaningful relationships from the overlapped spectra. Many commonly employed multivariate analysis techniques require the identity of the emission spectra of each component to be known or pure component pixels within the image, a condition rarely met in biological samples. Multivariate curve resolution (MCR) has proven extremely useful for analyzing hyperspectral and multispectral images of biological specimens because it can operate with little or no a priori information about the emitting species, making it appropriate for interrogating samples containing autofluorescence and unanticipated contaminating fluorescence. To demonstrate the unique ability of our hyperspectral imaging system coupled with MCR analysis techniques we will analyze hyperspectral images of four-color in-situ hybridized rat brain tissue containing 455 spectral pixels from 550 - 850 nm. Even though there were only four colors imparted onto the tissue in this case, analysis revealed seven fluorescent species, including contributions from cellular autofluorescence and the tissue mounting media. Spectral image analysis will be presented along with a detailed discussion of the origin of the fluorescence and specific illustrations of the adverse effects of ignoring these additional fluorescent species in a traditional microscopy experiment and a hyperspectral imaging system.

Compressive coded hyper spectral (HS) imaging actualizes compressed sampling and snapshot acquisition of HS data, whereas current recovery algorithms take too long time to make real-time HS imaging satisfactory. This paper proposes a deep learning approach for compressive HS imaging to shorten the recovery time. A fully-connected network is designed to train a block-based non-linear reconstruction operator. There is a mergence after obtaining the recovery 3D blocks, followed with a block edge mean filter. The contribution of this approach is that it uses deep neural network to do the reconstruction of the HS data for the first time and it has low-complexity and needs less memory because of operating on local patches. The proposed method was validated on a public available HS dataset and the experimental results show that this approach is superior to the state-of-the-art in the recovery accuracy, and dramatically improves the reconstruction speed by 400 ∼ 760 times.

Rooibos (Aspalathus linearis) and Honeybush (Cyclopia species) are teas indigenous to South Africa that are popularly consumed locally as well as internationally for its purported health benefits. Studies have confirmed anti-oxidant, antimutagenic and anticancer activity mainly due to the presence of polyphenols. Hyperspectral imaging (HSI) integrates conventional spectroscopy and imaging to obtain spectral and spatial information from a sample. The potential of short wave infrared (SWIR) hyperspectral imaging in combination with chemometric data analysis as a rapid quality control method for commercially important herbal teas was investigated. Images were acquired with the sisuChema short wave infrared (SWIR) hyperspectral pushbroom imaging system at a spectral range of 920 – 2514nm. Principal Component Analysis (PCA) was performed to remove background pixels and to visualize the data. Classes (i.e. Rooibos or Honeybush) were interactively assigned using the score images and plots. Partial Least Squares Discriminant Analysis (PLS-DA) classification models developed were subsequently used to accurately predict the species identity of commercially available herbal tea samples introduced as an external dataset, producing a visually interpretable result. The results showed that the labeling information is correct but that Honeybush tea is in most cases present in very low amounts. It is evident that hyperspectral imaging is an objective and non-destructive quality control method that can be successfully used to determine the botanical species included in herbal teas in addition to giving an indication of the abundance of each species in polyherbal tea mixtures.

A precision weed control system was developed to autonomously detect, identify and map weed species in the seedline of directly-seeded processing tomatoes and to apply a precise and lethal spray to the weed foliage. The weed sensor, based upon a hyperspectral imaging subsystem equipped with a line imaging spectrograph, had a spectral range of 385 - 810 nm at a 1.6 nm resolution and a spatial resolution of 0.4 mm across the seedline. Hyperspectral field images were collected in a processing tomato field in real-time from a continuously moving cultivation sled. A multivariate Bayesian classifier was constructed off-line to identify the plant species of each pixel in the hyperspectral images. At the pixel level, cross-validation results demonstrated a significant improvement in classifier performance on field data when site-specific classifier training was conducted. A site-specific classifier correctly recognized 95% of tomato foliage and more than 84%, of four weed species (black nightshade (Solanum nigrum L.), lambsquarter (Chenopodium album L.), red-root pigweed (Amaranthus retroflexus L.), and purslane (Portulaca oleracea L.)). At the plant level, nearly all plant species were correctly identified. The hyperspectral imaging system was robust in the identification of partially occluded foliage and provided an accurate weed map suitable for species-specific targeted herbicide application.

Hyperspectral (HS) image is a three dimensional data image where the 3rd dimension carries the wealth of spectrum information. HS image compression is one of the areas that has attracted increasing attention for big data processing and analysis. HS data has its own distinguishing feature which differs with video because without motion, also different with a still image because of redundancy along the wavelength axis. The prediction based method is playing an important role in the compression and research area. Reflectance distribution of HS based on our analysis indicates that there is some nonlinear relationship in intra-band. The Multilayer Propagation Neural Networks (MLPNN) with backpropagation training are particularly well suited for addressing the approximation function. In this paper, an MLPNN based predictive image compression method is presented. We propose a hybrid Adaptive Prediction Mechanism (APM) with MLPNN model (APM-MLPNN). MLPNN is trained to predict the succeeding bands by using current band information. The purpose is to explore whether MLPNN can provide better image compression results in HS images. Besides, it uses less computation cost than a deep learning model so we can easily validate the model. We encoded the weights vector and the bias vector of MLPNN as well as the residuals. That is the only few bytes it then sends to the decoder side. The decoder will reconstruct a band by using the same structure of the network. We call it an MLPNN decoder. The MLPNN decoder does not need to be trained as the weights and biases have already been transmitted. We can easily reconstruct the succeeding bands by the MLPNN decoder. APM constrained the correction offset between the succeeding band and the current spectral band in order to prevent HS image being affected by large predictive biases. The performance of the proposed algorithm is verified by several HS images from Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) reflectance dataset. MLPNN simulation results can improve prediction accuracy; reduce residual of intra-band with high compression ratio and relatively lower bitrates.

VTT Technical Research Centre of Finland has developed a Fabry-Perot Interferometer (FPI) based hyperspectral imager compatible with light weight UAV (Unmanned Aerial Vehicle) platforms (SPIE Proc. 7474¹, 8186B²). The FPI based hyperspectral imager was used in a UAV imaging campaign for forest and agriculture tests during the summer 2011 (SPIE Proc. 8174³). During these tests high spatial resolution Color-Infrared (CIR) images and hyperspectral images were recorded on separate flights. The spectral bands of the CIR camera were 500 - 580 nm for the green band, 580 - 700 nm for the red band and 700 - 1000 nm for the near infrared band. For the summer 2012 flight campaign a new hyperspectral imager is currently being developed. A custom made CIR camera will also be used. The system which includes both the high spatial resolution Color-Infrared camera and a light weight hyperspectral imager can provide all necessary data with just one UAV flight over the target area. The new UAV imaging system contains a 4 Megapixel CIR camera which is used for the generation of the digital surface models and CIR mosaics. The hyperspectral data can be recorded in the wavelength range 500 - 900 nm at a resolution of 10 - 30 nm at FWHM. The resolution can be selected from approximate values of 10, 15, 20 or 30 nm at FWHM.

The primary application of hyperspectral imaging is to classify/quantify objects/materials in the scene based on their spectral signatures. The spectral features that are useful can sometimes fall outside the spectral range of a single hyperspectral imager which is usually limited by the spectral response range of the sensor material of the focal plane array within the imager. For these wide spectrum applications, some users are combining data from two (or more) hyperspectral imaging systems. Aside from the optical alignment, size and synchronization issues involved in such a setup, the process of pixel co-registration, i.e. geometrically transforming data from the 2 hyperspectral imaging systems to overlay one another is tedious and complex. Headwall Photonics has integrated two of their off-the-shelf highly optimized hyperspectral imagers (Vis-NIR and SWIR) in an optically co-boresighted configuration together with a high performance data processor to produce a compact system which is easy to use and outputs wide spectrum pixel co-registered hyperspectral data. The process of pixel co-registration in this system is computationally very cheap enabling real-time wide-spectrum hyperspectral imaging applications. This paper presents actual imaging and performance data from these systems showing excellent pixel co-registration, sensitivity and spectral resolution.

Polarized spectra–hyperspectral data fusion technique was used to estimate the soluble sugar (SS), total nitrogen (N), and their ratio (SS/N), of greenhouse tomato leaves. Fresh tomato leaves of five different growth stages (seedling, flowering, initial fruiting, mid-fruiting and picking stage) and five different nitrogen treatments (severe stress 25%, moderate stress 50%, mild stress 75%, normal 100%, and excess 150%) at every stage were collected for spectra acquisition and SS and N determination. Polarized reflectance spectra were acquired with a polarization reflectance spectrum spectro-goniophotometer system and four polarization degree features were extracted. Hyperspectral data were collected with a hyperspectral imaging system and four reflectance spectrum features and eight image features were extracted. Initially, models were built with polarization degree features, image features, and spectral features respectively. Linear and nonlinear fusion methods were comparatively used for modeling based on normalized data of the three sources. The results suggest that the performances of SS/N models are better than those of N and SS models, and the prediction capability of the Support Vector Machine (SVM) models of N and SS/N are superior to those obtained with single kind feature. This work indicates that the polarized spectrum-hyperspectral multidimensional information detecting method can feasibly judge the tomato nutrient stress conditions. Multi-features data fusion analysis technique can enhance the prediction accuracy of spectral diagnostics technology in precision agriculture. Keywords: polarized spectra, hyperspectral, soluble sugar (SS), total nitrogen (N), data fusion, tomato leaf DOI: 10.25165/j.ijabe.20201302.4280 Citation: Zhu W J, Li J Y, Li L, Wang A C, Wei X H, Mao H P. Nondestructive diagnostics of soluble sugar, total nitrogen and their ratio of tomato leaves in greenhouse by polarized spectra–hyperspectral data fusion. Int J Agric & Biol Eng, 2020; 13(2): 189–197.

In the field of agro-business technology, computerization contributes to productivity, monetary turnover of events along local viability. The interest in tariffs in addition to the consistency analysis is influenced by the mix of leafy foods. The most tangible aspect of the food derived from the earth is the implementation that influences the need for, the customer’s desires as well as the judgment of the market. Although people may plan and assess, time-concentrated, complex, subjective, costly, and handily influenced by environmental variables is problematic. Subsequently, a shrewd natural product evaluation system is needed. Deep learning has achieved remarkable milestones in the field of conventional computers. In this article, we use deep learning techniques on the topic of hyperspectral image exploration. Unlike traditional machine vision exercises, the only thing to do with a gander is the spatial setting; our proposed solution would use both the spatial setting and the phantom relationship to enhance the hyperspectral image grouping. In clear words, we endorse four new deep learning models, in particular the 3D Convolutionary Neural Network (3D-CNN) and the Repetitive 3D Convolutionary Neural Network (R-3D-CNN) for hyperspectral image recognition.

Investigating the detection of peanuts in chopped nut products using hyperspectral imaging systems

In this paper, a comprehensive spatio-spectral and temporal analysis for Chromobacterium violaceum colonies is reported. A hyperspectral imaging (HSI) system is used to recover the spectral signatures of pigment production in a non-homogeneous media with high spectral resolution and high sensitivity in vivo, without destructing the sample. This non-contact sensing technique opens avenues to study the temporal growing of a specific section in the bacterial colony. Further, from a 580 [nm] and 764 [nm] spatio-spectral time series, a wild-type and mutant Chromobacterium violaceum strains are characterized. Such study provides quantitative information about kinetic parameters of pigment production and bacterial growing.

Pomological traits are the major factors determining the quality and price of fresh fruits. This research was aimed to investigate the feasibility of using two hyperspectral imaging (HSI) systems in the wavelength regions comprising visible to near infrared (VisNIR) (400−1000 nm) and short-wave infrared (SWIR) (935−1720 nm) for predicting four strawberry quality attributes (firmness—FF, total soluble solid content—TSS, titratable acidity—TA, and dry matter—DM). Prediction models were developed based on artificial neural networks (ANN). The entire strawberry VisNIR reflectance spectra resulted in accurate predictions of TSS (R2 = 0.959), DM (R2 = 0.947), and TA (R2 = 0.877), whereas good prediction was observed for FF (R2 = 0.808). As for models from the SWIR system, good correlations were found between each of the physicochemical indices and the spectral information (R2 = 0.924 for DM; R2 = 0.898 for TSS; R2 = 0.953 for TA; R2 = 0.820 for FF). Finally, data fusion demonstrated a higher ability to predict fruit internal quality (R2 = 0.942 for DM; R2 = 0. 981 for TSS; R2 = 0.976 for TA; R2 = 0.951 for FF). The results confirmed the potential of these two HSI systems as a rapid and nondestructive tool for evaluating fruit quality and enhancing the product’s marketability.

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.