Ground-based Hyperspectral Imaging Research Articles

Spatial Co-Registration and Spectral Concatenation of Panoramic Ground-Based Hyperspectral Images

Spatial Co-Registration and Spectral Concatenation of Panoramic Ground-Based Hyperspectral Images

Assessing the Impact of Spatial Resolution on the Estimation of Leaf Nitrogen Concentration Over the Full Season of Paddy Rice Using Near-Surface Imaging Spectroscopy Data.

Timely monitoring nitrogen status of rice crops with remote sensing can help us optimize nitrogen fertilizer management and reduce environmental pollution. Recently, the use of near-surface imaging spectroscopy is emerging as a promising technology that can collect hyperspectral images with spatial resolutions ranging from millimeters to decimeters. The spatial resolution is crucial for the efficiency in the image sampling across rice plants and the separation of leaf signals from the background. However, the optimal spatial resolution of such images for monitoring the leaf nitrogen concentration (LNC) in rice crops remains unclear. To assess the impact of spatial resolution on the estimation of rice LNC, we collected ground-based hyperspectral images throughout the entire growing season over 2 consecutive years and generated ten sets of images with spatial resolutions ranging from 1.3 to 450 mm. These images were used to determine the sensitivity of LNC prediction to spatial resolution with three groups of vegetation indices (VIs) and two multivariate methods Gaussian Process regression (GPR) and Partial least squares regression (PLSR). The reflectance spectra of sunlit-, shaded-, and all-leaf leaf pixels separated from background pixels at each spatial resolution were used to predict LNC with VIs, GPR and PLSR, respectively. The results demonstrated all-leaf pixels generally exhibited more stable performance than sunlit- and shaded-leaf pixels regardless of estimation approaches. The predictions of LNC required stage-specific LNC~VI models for each vegetative stage but could be performed with a single model for all the reproductive stages. Specifically, most VIs achieved stable performances from all the resolutions finer than 14 mm for the early tillering stage but from all the resolutions finer than 56 mm for the other stages. In contrast, the global models for the prediction of LNC across the entire growing season were successfully established with the approaches of GPR or PLSR. In particular, GPR generally exhibited the best prediction of LNC with the optimal spatial resolution being found at 28 mm. These findings represent significant advances in the application of ground-based imaging spectroscopy as a promising approach to crop monitoring and understanding the effects of spatial resolution on the estimation of rice LNC.

Ground-based hyperspectral imaging as a tool to identify different carbonate phases in natural cliffs

ABSTRACTRecent research has shown hyperspectral imaging to be a powerful tool to distinguish carbonate phases with slight compositional differences on quarry cliff faces. The traditional remote sensing set-up uses an optimal short distance between the hyperspectral camera mounted on a tripod and a quarry wall characterized by a planar, mostly unweathered surface. Here we present results of a modified workflow geared to the application of ground-based hyperspectral imaging of rough and weathered cliff faces in order to map large scale dolomite bodies from a distance of up to several kilometres. The goal of the study was to determine unique spectral properties of fracture-controlled dolomite bodies in order to be able to distinguish them from a dolomitic host rock. In addition, the impact of weathering on carbonate phases and thus, the modification of the spectral signature between altered and unaltered carbonates is assessed. The spectral analysis is complemented by ICP-AES (inductively coupled plasma atomic emission spectroscopy) measurements of the spectrally measured powders. Furthermore, we examined the detection limits and characterisation potential of dolomite bodies from hyperspectral images captured at varying distances from cliff faces in the study area. Hyperspectral images of 10 natural cliffs distributed across the Central Oman Mountains were obtained with a Push broom scanner system. The high resolution of 5.45 nm (288 bands in total) enabled the visualization of small-scale changes in the near infrared continuous spectrum of all present lithofacies types. The determination of dolomite bodies of varying sizes (metre to hundreds of metres) on natural cliffs was achieved with the hyperspectral mapping approach and mapping results have been tested with the position of visually defined dolomite bodies on field panoramas. Spectra of natural cliffs contain a strong absorption peak indicative for iron which is absent in spectra of unweathered sample powders. However, ICP-AES analysis of powders revealed relatively low contents of iron of 12,392 ppm. The strong peaks in field images are interpreted as linked to intensive weathering associated with the precepitation of goethite, hematite, specularite and manganese as well as intensive dedolomitization. Dedolomitization is indicated by calcitic spectra derived from the dolomite bodies. The spectral difference of laboratory and field spectra interferes significantly the application of laboratory spectra of powdered samples for the identification of dolomite bodies in the field. Furthermore, the process of late dedolomitization puts an additional challenge on the determination of dolomite bodies. Due to these strong spectral variations between laboratory and field spectra, we recommend that the mapping approaches should not solely rely on spectral algorithms but also consider normal light field panoramas and representative outcrop analysis. We also note that the quality of resolution is too low for the determination of small-scale variations of diagenetic phases at distances larger than 4 km. However, when the limitations mentioned are taken into account, hyperspectral imaging proves to be a powerful tool that helps in the determination of the distribution of diagenetic phases, even in challenging conditions.

Evaluating laboratory-based classification potentials of heavy metal contaminated soils using spectro-radiometer and hyper-spectral imagery

In this study, we constructed indoor testbed to apply aerial hyper-spectral imaging sensor was to monitoring to heavy metal contaminated in soil. And a ground-based spectro-radiometer data and hyper-spectral image data were obtained for soil samples artificially contaminated with two heavy metals (Cu, Pb). Based on the comparison of spectro-radiometer and actual concentration data, specific wavelength band was classified by linear regression analysis and statistical analysis. We have established a preprocessing methodology that represents the most correlated specific wavelengths through logarithmic and derivative transformations on the reflectance data for each concentration gradient of heavy metals. Among the various methods, (Log (1/R))′ conversion was confirmed as the most effective pre-processing method for extracting specific wavelength. Then, the spectral characteristic patterns of the soil contaminated with heavy metals were confirmed through the hyper-spectral image data constructed in the indoor test bed, and spectral angle mapper which is one of the spectral characteristic matching methods was applied to confirm the classification feasibility of soil contaminated with the heavy metals.

Radiometric Correction and 3D Integration of Long-Range Ground-Based Hyperspectral Imagery for Mineral Exploration of Vertical Outcrops

Recently, ground-based hyperspectral imaging has come to the fore, supporting the arduous task of mapping near-vertical, difficult-to-access geological outcrops. The application of outcrop sensing within a range of one to several hundred metres, including geometric corrections and integration with accurate terrestrial laser scanning models, is already developing rapidly. However, there are few studies dealing with ground-based imaging of distant targets (i.e., in the range of several kilometres) such as mountain ridges, cliffs, and pit walls. In particular, the extreme influence of atmospheric effects and topography-induced illumination differences have remained an unmet challenge on the spectral data. These effects cannot be corrected by means of common correction tools for nadir satellite or airborne data. Thus, this article presents an adapted workflow to overcome the challenges of long-range outcrop sensing, including straightforward atmospheric and topographic corrections. Using two datasets with different characteristics, we demonstrate the application of the workflow and highlight the importance of the presented corrections for a reliable geological interpretation. The achieved spectral mapping products are integrated with 3D photogrammetric data to create large-scale now-called “hyperclouds”, i.e., geometrically correct representations of the hyperspectral datacube. The presented workflow opens up a new range of application possibilities of hyperspectral imagery by significantly enlarging the scale of ground-based measurements.

Estimation of rice protein content before harvest using ground-based hyperspectral imaging and region of interest analysis

Protein content, which represents rice taste quality, must be estimated in order to create a harvesting plan as well as next year’s basal dressing fertilizer application plan. Ground-based hyperspectral imaging with high resolution (1 × 1 mm per pixel) was used for estimating the protein content of brown rice before harvest. This paper compares the estimation accuracy of rice protein content estimation models generated from the mean reflectances of five regions of interest (ROIs): the overall target area, dark area (less illuminated parts of the rice plants), canopy area (leaves, yellow leaves, and ears), leaf area, and ear and yellow leaf area. The size of the target sampling area was 0.85 × 0.85 m. An R + G + B histogram and a GNDVI–NDVI image were used to separate the target area into the individual ROIs. The values of the coefficient of determination R 2 and the root mean square error of prediction (RMSE) were similar for each model: R 2 ranged from 0.83 to 0.86 and RMSE ranged from 0.27 to 0.30% for all models except for the dark area model, where R 2 = 0.76 and RMSE = 0.35%. There were no significant differences in the magnitude of the estimation error among all models. This result indicates that it is not necessary to obtain an image with a ground resolution that is greater than 0.85 × 0.85 m per pixel to estimate rice protein content before harvest. This result should provide useful information when deciding the altitude of platforms for imaging rice fields.

Integrated ground-based hyperspectral imaging and geochemical study of the Eagle Ford Group in West Texas

This study used ground-based hyperspectral imaging to map an outcrop of the Eagle Ford Group in west Texas. The Eagle Ford Group consists of alternating layers of mudstone – wackestone, grainstone – packstone facies and volcanic ash deposits with high total organic content deposited during the Cenomanian – Turonian time period. It is one of the few unconventional source rock and reservoirs that have surface representations.Ground-based hyperspectral imaging scanned an outcrop and hand samples at close ranges with very fine spatial resolution (centimeter to sub-millimeter). Spectral absorption modeling of clay minerals and calcite with the modified Gaussian model (MGM) allowed quantification of variations of mineral abundances. Petrographic analysis confirmed mineral identifications and shed light on sedimentary textures, and major element geochemistry supported the mineral quantification. Mineral quantification resulted in mapping of mudstone – wackestone, grainstone – packstone facies and bentonites (volcanic ash beds). The lack of spatial associations between the grainstones and bentonites on the outcrop calls into question the hypothesis that the primary productivity is controlled by iron availability from volcanic ash beds. Enrichment of molybdenum (Mo) and uranium (U) indicated “unrestricted marine” paleo-hydrogeology and anoxic to euxinic paleo-redox bottom water conditions.Hyperspectral remote sensing data also helped in creating a virtual outcrop model with detailed mineralogical compositions, and provided reservoir analog to extract compositional and geo-mechanical characteristics and variations. The utilization of these new techniques in geo-statistical analysis provides a workflow for employing remote sensing in resource exploration and exploitation.

Ground-based hyperspectral imaging and terrestrial laser scanning for fracture characterization in the Mississippian Boone Formation

Petroleum geoscientists have been using cores and well logs to study source rocks and reservoirs, however, the inherent discontinuous nature of these data cannot account for horizontal heterogeneities. Modern exploitation requires better understanding of important source rocks and reservoirs at outcrop scale. Remote sensing of outcrops is becoming a first order tool for reservoir analog studies including horizontal heterogeneities. This work used ground-based hyperspectral imaging, terrestrial laser scanning (TLS), and high-resolution photography to study a roadcut of the Boone Formation at Bella Vista, northwest Arkansas, and developed an outcrop model for reservoir analog analyses.The petroliferous Boone Formation consists of fossiliferous limestones interbedded with chert of early Mississippian age. We used remote sensing techniques to identify rock types and to collect 3D geometrical data. Mixture tuned matched filtering classification of hyperspectral data show that the outcrop is mostly limestones with interbedded chert nodules. 1315 fractures were classified according to their strata-bounding relationships, among these, larger fractures are dominantly striking in ENE – WSW directions. Fracture extraction data show that chert holds more fractures than limestones, and both vertical and horizontal heterogeneities exist in chert nodule distribution.Utilizing ground-based remote sensing, we have assembled a virtual outcrop model to extract mineral composition as well as fracture data from the model. We inferred anisotropy in vertical fracture permeability based on the dominancy of fracture orientations, the preferential distribution of fractures and distribution of chert nodules. These data are beneficial in reservoir analogs to study rock mechanics and fluid flow, and to improve well performances.

Comparison and selection of vegetation indices for detection of Sclerotinia Stem Rot on oilseed rape leaves using ground-based hyperspectral imaging

A ground-based hyperspectral imaging system covering the spectral range of 384–1034 nm was used for Sclerotinia Stem Rot (SSR) detection. Two sample sets of oilseed leaves were collected. Four vegetation indices were extracted and evaluated by analysis of variance (ANOVA) combined with linear discriminant analysis (LDA) for the two sample sets. Discriminant models were built using the 4 vegetation indices. The discriminant results of the two sample sets were good with classification accuracies of the calibration set and the prediction set over 85%. The overall results indicated that vegetation indices calculated from ground-based hyperspectral imaging could be used as reliable and accurate indices for SSR detection.

Ground-Based Hyperspectral Image Analysis of the Lower Mississippian (Osagean) Reeds Spring Formation Rocks in Southwestern Missouri

Ground-based hyperspectral imaging is fairly new for studying near-vertical rock exposures where airborne or satellite-based imaging fail to provide useful information. In this study, ground-based hyperspectral image analysis was performed on a roadcut, where diagenetic tripolite facies is observed in southwestern Missouri. Laboratory-based reflectance spectroscopy and hyperspectral image analyses were also performed on collected samples. Image classification was performed using Spectral Feature Fitting (SFF) and Mixture-tuned Match Filtering (MTMF) with laboratory- and image-derived end-member spectra. SFF provided thorough yet detailed classification, whereas MTMF provided information on the relative abundances of the lithologies. Ground-based hyperspectral imaging demonstrated its potential to aid geological studies providing valuable information on mineralogical and lithological variations rapidly and with two-dimensional continuity in inaccessible rock faces of near-vertical outcrops. The results showed decreasing tripolite abundance going downward in the investigated vertical succession. Also, a leaching pattern has been observed such that persistent and continuous limestone layers become lenses and patches towards the upper portion of the outcrop. These observations show that the effect of tripolitization decreases when going deeper in the succession, suggesting that the fluid responsible for the weathering of siliceous precursors may have been flowing from top to bottom and thus have had a meteoric origin.

Study of Upper Albian rudist buildups in the Edwards Formation using ground-based hyperspectral imaging and terrestrial laser scanning

Ground-based hyperspectral imaging is used for development of digital outcrop models which can facilitate detailed qualitative and quantitative sedimentological analysis and augment the study of depositional environment, diagenetic processes, and hydrocarbon reservoir characterization in areas which are physically inaccessible. For this investigation, ground-based hyperspectral imaging is combined with terrestrial laser scanning to produce mineralogical maps of Late Albian rudist buildups of the Edwards formation in the Lake Georgetown Spillway in Williamson County, Texas. The Edwards Formation consists of shallow water deposits of reef and associated interreef facies. It is an aquifer in western Texas and was investigated as a hydrocarbon play in south Texas. Hyperspectral data were registered to a geometrically accurate laser point cloud-generated mesh with sub-pixel accuracy and were used to map compositional variation by distinguishing spectral properties unique to each material. More calcitic flat-topped toucasid-rich bioherm facies were distinguished from overlying porous sucrosic dolostones, and peloid wackestones and packstones of back-reef facies. Ground truth was established by petrographic study of samples from this area. This research integrates high-resolution datasets to analyze geometrical and compositional properties of this carbonate formation at a finer scale than traditional methods have achieved and to model the geometry and composition of rudist buildups.

Remote detection of fluid-related diagenetic mineralogical variations in the Wingate Sandstone at different spatial and spectral resolutions

Well-exposed eolian units of the Jurassic system on the Colorado Plateau including the Wingate Sandstone, show prominent color variations throughout southeastern Utah due to diagenetic changes that include precipitation and/or removal of iron oxide, clay, and carbonate cement. Spatially variable characteristic diagenetic changes suggest fluid-rock interactions through the sandstone. Distinctive spectral signatures of diagenetic minerals can be used to map diagenetic mineral variability and possibly fluid-flow pathways. The main objective of this work was to identify characteristic diagenetic minerals, and map their spatial variability from regional to outcrop scale in Wingate Sandstone exposures of Lisbon Valley, Utah. Laboratory reflectance spectroscopy analysis of the samples facilitated identification of diagnostic spectral characteristics of the common diagenetic minerals and their relative abundances between altered and unaltered Wingate Sandstone. Comparison of reflectance spectroscopy with satellite, airborne, and ground-based imaging spectroscopy data provided a method for mapping and evaluating spatial variations of diagenetic minerals. The Feature-oriented Principal Component Selection method was used on Advanced Spaceborne Thermal Emission and Reflection Radiometer data so as to map common mineral groups throughout the broader Wingate Sandstone exposure in the area. The Minimum Noise Fraction and Spectral Angle Mapper methods were applied on airborne HyMap and ground-based hyperspectral imaging data to identify and map mineralogical changes. The satellite and airborne data showed that out of 25.55km2 total exposure of Wingate Sandstone in Lisbon Valley, unaltered sandstone cover 12.55km2, and altered sandstone cover 8.90km2 in the northwest flank and 5.09km2 in the southern flank of the anticline. The ground-based hyperspectral data demonstrated the ability to identify and map mineral assemblages with two-dimensional lateral continuity on near-vertical rock faces. The results showed that 39.71% of the scanned outcrop is bleached and 20.60% is unbleached while 6.33% remain unclassified, and 33.36% is masked-out as vegetation. The bleached and unbleached areas are alternating throughout the vertical face of the outcrop. The relative hematite abundance observed in the unbleached areas are somewhat symmetrical. This indicates fairly similar reaction intensities along the upper and lower reaction fronts observed in the vertical section. The distribution geometry and relative abundances of diagenetic minerals not only suggest multiple paths of fluid-flow in Wingate Sandstone but also provides some insight about relative direction of past fluid-flow.

Nitrogen prediction model of rice plant at panicle initiation stage using ground-based hyperspectral imaging: growing degree-days integrated model

Ground-based hyperspectral imaging was applied to rice plants at the panicle initiation stage to estimate nitrogen content. We developed a partial least squares regression (PLSR) model that incorporated both the reflectance and growing degree-days (GDD) to account for differences in growing temperature conditions across a 3-year period. The acquired images were divided into two components: (1) the rice plant and (2) other elements (e.g., irrigation water, soil background) by using the GreenNDVI − NDVI equation. Rice plant reflectance (Ref RICE ) was calculated as the ratio of rice plant reflectance to that of a reference board. Three types of PLSR models were constructed: 1-year, 2-year, and 2-year GDD. Mutual estimation was used to infer the predictive power of the three models, which was calculated by estimating the values for the other years. The root mean square error of prediction (RMSE) of the mutual estimation for the 1- and 2-year PLSR models was high because of overestimation and underestimation. In contrast, the RMSE of the mutual estimation for the 2-year GDD PLSR models clearly decreased. It was inferred that hyperspectral imaging at 400–1000 nm could not predict variation in the amount of growth caused by weather variation expressed as GDD. This study indicates that the combination of reflectance and temperature data could be used to potentially construct an adaptable model to identify variance in growing conditions.

Deblurring and Sparse Unmixing of Hyperspectral Images Using Multiple Point Spread Functions

This paper is concerned with deblurring and spectral analysis of ground-based astronomical images of space objects. A numerical approach is provided for deblurring and sparse unmixing of ground-based hyperspectral images (HSIs) of objects taken through atmospheric turbulence. Hyperspectral imaging systems capture a three-dimensional (3D) datacube (tensor) containing two-dimensional (2D) spatial information and one-dimensional (1D) spectral information at each spatial location. Pixel intensities vary with wavelength bands, providing a spectral trace of intensity values and generating a spatial map of spectral variation (spectral signatures of materials). The deblurring and spectral unmixing problem is quite challenging since the point spread function (PSF) depends on the imaging system as well as the seeing conditions and is wavelength varying. We show how to efficiently construct an optimal Kronecker product--based preconditioner, and provide numerical methods for estimating the multiple PSFs using spectral data from an isolated (guide) star for joint deblurring and sparse unmixing of HSI datasets in order to spectrally analyze the image objects. The methods are illustrated with numerical experiments on a commonly used test example, a simulated HSI of the Hubble Space Telescope satellite.

Hyperspectral detection of a subsurface CO2 leak in the presence of water stressed vegetation.

Remote sensing of vegetation stress has been posed as a possible large area monitoring tool for surface CO2 leakage from geologic carbon sequestration (GCS) sites since vegetation is adversely affected by elevated CO2 levels in soil. However, the extent to which remote sensing could be used for CO2 leak detection depends on the spectral separability of the plant stress signal caused by various factors, including elevated soil CO2 and water stress. This distinction is crucial to determining the seasonality and appropriateness of remote GCS site monitoring. A greenhouse experiment tested the degree to which plants stressed by elevated soil CO2 could be distinguished from plants that were water stressed. A randomized block design assigned Alfalfa plants (Medicago sativa) to one of four possible treatment groups: 1) a CO2 injection group; 2) a water stress group; 3) an interaction group that was subjected to both water stress and CO2 injection; or 4) a group that received adequate water and no CO2 injection. Single date classification trees were developed to identify individual spectral bands that were significant in distinguishing between CO2 and water stress agents, in addition to a random forest classifier that was used to further understand and validate predictive accuracies. Overall peak classification accuracy was 90% (Kappa of 0.87) for the classification tree analysis and 83% (Kappa of 0.77) for the random forest classifier, demonstrating that vegetation stressed from an underground CO2 leak could be accurately discerned from healthy vegetation and areas of co-occurring water stressed vegetation at certain times. Plants appear to hit a stress threshold, however, that would render detection of a CO2 leak unlikely during severe drought conditions. Our findings suggest that early detection of a CO2 leak with an aerial or ground-based hyperspectral imaging system is possible and could be an important GCS monitoring tool.

Integrate Growing Temperature to Estimate the Nitrogen Content of Rice Plants at the Heading Stage Using Hyperspectral Imagery

Ground-based hyperspectral imaging was used for estimating the nitrogen content of rice plants at the heading stage. The images were separated into two parts: 1) the rice plant; and 2) other elements using the equation of “GreenNDVI-NDVI.” was calculated as the ratio of the reflectance of the rice plant to that of a reference board. Partial least square (PLS) model using reflectance data (R-PLS model) and PLS model using reflectance and temperature data (RT-PLS) was constructed to compare the accuracy between them. RT-PLS model was developed to improve the accuracy of R-PLS model by considering the differences of weather condition among years. When the precision (R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) and accuracy [root-mean-square error (RMSE) and relative error (RE)] of each R-PLS model were evaluated for each year using twofold cross-validation, R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ranged from 0.42 to 0.81, RMSE ranged from 0.81 to 1.13 gm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> , and RE ranged from 10.1% to 11.8%. When R-PLS model of each year was used to predict the other years' data to determine the predictive power, RMSE values were higher (ranging from 1.40 to 5.82 gm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> ) than those in each year's validation value due to over- or underestimation. When an R-PLS model based on the data of 2 years was fitted, RMSE ranged from 1.11 to 4.15 gm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> and RE ranged from 13.7% to 42.8%. By contrast, in RT-PLS models, RMSE and RE fell to less than 1.21 gm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> and 12.3%, respectively. Thus, a combination of reflectance and temperature data was useful for constructing a model of rice plant at the heading stage.

Retrieval of cirrus optical thickness and assessment of ice crystal shape from ground-based imaging spectrometry

Abstract. A ground-based hyperspectral imaging spectrometer (AisaEAGLE, manufactured by Specim Ltd., Finland) is applied to measure downward spectral radiance fields with high spatial (1024 spatial pixels within 36.7° field of view), spectral (488 spectral pixels, 400–970 nm, 1.25 nm full width at half maximum), and temporal (4–30 Hz) resolution. The calibration, measurement and data evaluation procedures are introduced. A new method is presented to retrieve the cirrus optical thickness (τci) using the spectral radiance data collected by AisaEAGLE. The data were collected during the Cloud Aerosol Radiation and tuRbulence of trade wInd cumuli over BArbados (CARRIBA) project in 2011. The spatial inhomogeneity of the investigated cirrus is characterised by the standard deviation of the retrieved τci as well as the width of its frequency distribution. By comparing measured and simulated downward solar spectral radiance as a function of scattering angle, some evidence of the prevailing cirrus ice crystal shape can be obtained and subsequently used to substantiate the retrieval of τci. The sensitivity of the retrieval method with respect to surface albedo, effective radius (reff), cloud height and ice crystal shape is quantified. An enhanced sensitivity of the retrieved τci is found with respect to the surface albedo (up to 30%) and ice crystal shape (up to 90%). The sensitivity with regard to the effective ice crystal radius (&amp;amp;leq; 5%) and the cloud height (&amp;amp;leq; 0.5%) is rather small and can be neglected.

Influence of solar irradiance on hyperspectral imaging-based plant recognition for autonomous weed control

Canopy reflectance in the visible and near infrared region (384–810 nm) was examined to discriminate between plant species, grown under various sunlight intensities, using ground-based hyperspectral imaging technology. Black nightshade and pigweed were grown with processing tomatoes under two levels of solar irradiance. The canonical Bayesian classifiers based on the full spectral range (400–795 nm) achieved an overall classification accuracy of 88.2% for the low solar irradiance treatment and 95.1% for the high solar irradiance treatment, using internal cross-validation analysis. The plant species exposed to higher solar irradiance were more easily distinguished in the feature space. The classifier trained with the plants grown in the low solar irradiance treatment was more robust to varying sunlight conditions; and correctly identified 88.7% of the plants for both sunlight intensity growing conditions. Global calibration achieved an optimum classification rate of 90% over the studied range of solar irradiance using 270 plants in the global domain as the training samples. It provided an alternative method to mitigate the effect on species discrimination and to improve classification robustness due to changing canopy optical properties across variation in solar irradiance during plant establishment.

In situ estimation of water quality parameters in freshwater aquaculture ponds using hyperspectral imaging system

Knowledge of water quality parameters is integral to sustainability of freshwater aquaculture operations that raise ornamental fish. Our objective in this study is to evaluate the ability of a mobile, ground-based hyperspectral (HS) imaging sensor to determine chlorophyll-a (Chl-a) concentrations in working aquaculture ponds, which represent manipulated, shallow, nutrient-rich systems, and to determine the effect of using submerged reflectance targets on the accuracy of Chl-a estimation. We collected Chl-a measurements from aquaculture ponds ranging from 0.8 to 494 μ g / L . Chl-a measurements showed a strong correlation with two-band and three-band spectral indices computed from the HS image reflectance. Coefficient of determination ( R 2 ) values of 0.975 and 0.982 were obtained for the two- and three-band models, respectively, using spectra captured from the submerged target at 10 cm depth. Using spectra captured from water (no submerged targets), R 2 values were slightly lower at 0.833 and 0.862 for two- and three-band models. Data from the submerged target at 30 cm depth had the lowest correlation with measured chlorophyll-a concentrations, potentially due to variations in water column properties and shadows cast by the platform. Modeling total Phosphorous (P) and Nitrogen (N) concentrations of the collected samples with the spectral indices sensitive to Chl-a concentrations showed a moderate level of correlation. Removing a model outlier (observation with maximum N and P concentrations) led to a significant increase in the models’ coefficient of determination ( e.g. from 0.478 to 0.823 for the P model using three-band index values), which highlighted the possibility of using HS imagery to estimate N and P concentrations and the need for more research to model the interrelationships between Chl-a and nutrient concentrations in aquaculture water systems.

An Adaptive Approach for the Progressive Integration of Spatial and Spectral Features When Training Ground-Based Hyperspectral Imaging Classifiers

The use of hyperspectrometers as analytical tools for determining surface material properties in ground-based applications introduces the need of integrating spatial and spectral hyperspectral cube components. A neural-network-based approach is presented in this paper with the aim of automatically adapting to the spatiospectral characteristics of samples in a problem domain so that the most efficient classification can be obtained. Its main application would be in inspection and quality control tasks. The system core is an Artificial Neural Network-based hyperspectral processing unit able to perform the online classification of the material based on the spatiospectral patterns provided by a set of pixels. A training adviser is implemented to automate the determination of the minimum spatial window size, as well as the optimum spectrospatial feature set leading to the desired classification in terms of the available ground truth. Several tests have been carried out on synthetic and real data sets. In particular, the proposed approach is used to discriminate samples of synthetic and real materials, where the pixel resolution implies that a material is characterized by spectral patterns of combinations of pixels.