High Resolution Hyperspectral Imagery Research Articles

Negative sea level anomalies with extreme low tides in the South-West Indian Ocean shape Reunion Island’s fringing coral reef flats

Among induced mass-mortality events on coral reef, extreme low tides may ultimately lead to considerable reef community deaths on intertidal reef flats due to unusually long and significant aerial exposure. Here, we report an extensive coral mortality event induced by a negative sea level anomaly (nSLA) that occurred across Reunion Island during the austral winter season between June and October 2015 preceding the 2015–2016 El Niño Southern Oscillation (ENSO) event. The nSLA was strong and long in duration with a rapid drop of 35 cm in the mean sea level over a one-month period. Surveys conducted over seven reef flat sites before and after the nSLA revealed that mean coral cover drastically decreased from 54.5 ± 12.7% in early 2015, to 27.4 ± 6.9% in November 2015, which is an equivalent cover loss of 50% following the 2015 nSLA event. The shallowest sites showed a greater decrease in coral cover while the deepest parts of the reef flat remained unaffected. We found a significant correlation between the bathymetry and the relative coral cover variation. Using this relationship between depth and coral cover changes, high-resolution hyperspectral imagery and Lidar bathymetric airborne data, we mapped the impacts of this event at the scale of the whole reef. Overall the modeled loss reached 13.0 ha, which represents a decrease of 45.5% of all live coral cover in this area during the 2015 nSLA event. The impact of a nSLA on emersion times is much greater than the regular variation in tide amplitude between neap and spring tides, reaching new bathymetric ranges that are usually stable in terms of water submersion. Temporal variation of coral cover on Reunion Island reef flat revealed regular decreases to be compared with mean low-water-level events among other sea and climatic related disturbances and stressors.

Unsupervised stepwise extraction of offshore aquaculture ponds using super-resolution hyperspectral images

Despite offshore aquaculture brings great economic benefits, it also has destructive effects on the ecological environment of coastal regions. Therefore, the accurate monitoring of offshore aquaculture areas is vital. Existing methods for extracting aquaculture ponds are still limited by the lack of high-resolution hyperspectral imagery and inadequate feature extraction. In this paper, we proposed an unsupervised aquaculture ponds extraction method based on hyperspectral imagery super-resolution, feature fusion and stepwise extraction strategy. First, the resolution of the original hyperspectral imagery is enhanced via a deep learning-based super-resolution method. Then we introduce a feature fusion method by combining multi-dimensional features including elevation, reflectance, biochemistry, principal component analysis, and fast-Fourier transform features to enhance the feature sensitivity to the aquaculture ponds. The aquaculture ponds are finally extracted via a step-wise extraction strategy. Experiments on ZH-1 and GF-7 imageries are conducted, and the experimental results show that our method achieves an OA of 97.9% on the aquaculture pond extraction and that the method can be generalized for the object extraction on multispectral datasets. Ablation experiments show that all three of our innovative modules can effectively improve the extraction accuracy and confirm the generalization of the method.

Investigating mixing patterns of suspended sediment in a river confluence using high-resolution hyperspectral imagery

In a river confluence, the fluvial process is complex because of the merging of different flows from rivers with distinct morphologies. However, the mixing of suspended sediment in river confluences has been inadequately investigated owing to the low resolution of conventional techniques for measuring suspended sediment. In this study, we investigated the mixing of suspended sediment at the confluence of the Hwang and Nakdong Rivers in South Korea by analyzing the spectral characteristics of suspended sediment from high-resolution hyperspectral data. We retrieved the suspended sediment concentration (SSC) from high-resolution hyperspectral imagery using the conversion technique, which is a cluster-based machine learning regression with optical variability. Hyperspectral clustering was first applied to classify the water regions of the main river (Nakdong) and its tributary (Hwang). The clustering result at a near-field after the confluence point was strongly related to the mixing degree of flow and sediments from two different rivers. Through segmentation, regressors with hyperspectral clusters enabled accurate SSC measurement. In particular, they precisely retrieved the concentrations adjacent to the mixing layer, even when the relationship between the spectral data and suspended sediment concentrations of the main river and tributary was substantially different. Using a detailed SSC map, we compared two cases of mixing patterns by calculating the variation in sediment concentration along the downstream river confluence. We found that the mixing pattern was different because of the flow velocity and a large sandbar near the stagnant area even though river discharge was similar for both cases. In the slow mixing case, the sandbar caused a velocity-deficit with a wake that reduced the variance of SSC due to the irregular flow near the mixing layer. In this case, dual counter-rotating secondary flow cells also limited mixing across the post-confluence reach. However, when the wake and dual counter-rotating secondary flows weakened due to the erosion of the sandbar and lower flow velocity, the transverse mixing coefficient increased by approximately 40 % compared to the slow mixing case. This study demonstrated that the hyperspectral approach can be used to investigate the mixing of suspended sediment in greater detail in complex river systems, as it provides high-resolution information that conventional measurement methods cannot explore.

Radiative transfer model inversion using high-resolution hyperspectral airborne imagery - Retrieving maize LAI to access biomass and grain yield.

Mapping crop within-field yield variability provide an essential piece of information for precision agriculture applications. Leaf Area Index (LAI) is an important parameter that describes maize growth, vegetation structure, light absorption and subsequently maize biomass and grain yield (GY). The main goal for this study was to estimate maize biomass and GY through LAI retrieved from hyperspectral aerial images using a PROSAIL model inversion and compare its performance with biomass and GY estimations through simple vegetation index approaches. This study was conducted in two separate maize fields of 12 and 20 ha located in north-west Mexico. Both fields were cultivated with the same hybrid. One field was irrigated by a linear pivot and the other by a furrow irrigation system. Ground LAI data were collected at different crop growth stages followed by maize biomass and GY at the harvesting time. Through a weekly/biweekly airborne flight campaign, a total of 19 mosaics were acquired between both fields with a micro-hyperspectral Vis-NIR imaging sensor ranging from 400 to 850 nanometres (nm) at different crop growth stages. The PROSAIL model was calibrated and validated for retrieving maize LAI by simulating maize canopy spectral reflectance based on crop-specific parameters. The model was used to retrieve LAI from both fields and to subsequently estimate maize biomass and GY. Additionally, different vegetation indices were calculated from the aerial images to also estimate maize yield and compare the indices with PROSAIL based estimations. The PROSAIL validation to retrieve LAI from hyperspectral imagery showed a R2 value of 0.5 against ground LAI with RMSE of 0.8 m2/m2. Maize biomass and GY estimation based on NDRE showed the highest accuracies, followed by retrieved LAI, GNDVI and NDVI with R2 value of 0.81, 0.73, 0.73 and 0.65 for biomass, and 0.83, 0.69, 0.73 and 0.62 for GY estimation, respectively. Furthermore, the late vegetative growth stage at V16 was found to be the best stage for maize yield prediction for all studied indices.

Residual Effect and N Fertilizer Rate Detection by High-Resolution VNIR-SWIR Hyperspectral Imagery and Solar-Induced Chlorophyll Fluorescence in Wheat

Adjusting nitrogen (N) fertilization and accounting for the legacy of past N fertilizer application (i.e., residual N) based on remote sensing estimation of crop nutritional status may increase resource efficiency and promote sustainable management of cropping systems. Our main goal was to evaluate the potential of hyperspectral airborne imagers and ground-level sensors for identifying N fertilizer rates and the residual N effect from the previous crop fertilization in a maize/wheat rotation. A two-season field trial that provided various combinations of N rates and residual N response was established in central Spain. Ground-level sensors and aerial hyperspectral images were used to calculate vegetation indices (VIs). In addition, the solar-induced chlorophyll fluorescence (SIF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">760</sub> ) was estimated by the Fraunhofer line-depth method using high-resolution hyperspectral imagery, and together with biophysical modeling, biochemical and biophysical constituents at canopy scales were retrieved. N uptake, N output, grain N concentration, and proximal sensors discriminated between different N fertilizer rates and identified the residual effect when it was relevant. Structural, photosynthetic pigments and short-wave infrared region (SWIR)-based VIs, together with SIF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">760</sub> and the chlorophyll <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$a + b$ </tex-math></inline-formula> ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$C_{{\mathrm {ab}}}$ </tex-math></inline-formula> ), biomass, and the leaf area index (LAI), performed similarly on N rate detection. However, the residual effect of nitrification inhibitors was only detected by the structural (NDVI and OSAVI), chlorophyll (CCCI and NDRE), blue/green, NIR-SWIR ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\text{N}_{850,1510}$ </tex-math></inline-formula> ) indices, SIF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">760</sub> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$C_{{\mathrm {ab}}}$ </tex-math></inline-formula> , biomass, and the LAI. This study confirmed the ability of remote sensing to identify N rates at early growth stages and highlighted its potential to detect residual N in crop rotation.

The Influence of Aerial Hyperspectral Image Processing Workflow on Nitrogen Uptake Prediction Accuracy in Maize

A meticulous image processing workflow is oftentimes required to derive quality image data from high-resolution, unmanned aerial systems. There are many subjective decisions to be made during image processing, but the effects of those decisions on prediction model accuracy have never been reported. This study introduced a framework for quantifying the effects of image processing methods on model accuracy. A demonstration of this framework was performed using high-resolution hyperspectral imagery (&lt;10 cm pixel size) for predicting maize nitrogen uptake in the early to mid-vegetative developmental stages (V6–V14). Two supervised regression learning estimators (Lasso and partial least squares) were trained to make predictions from hyperspectral imagery. Data for this use case were collected from three experiments over two years (2018–2019) in southern Minnesota, USA (four site-years). The image processing steps that were evaluated include (i) reflectance conversion, (ii) cropping, (iii) spectral clipping, (iv) spectral smoothing, (v) binning, and (vi) segmentation. In total, 648 image processing workflow scenarios were evaluated, and results were analyzed to understand the influence of each image processing step on the cross-validated root mean squared error (RMSE) of the estimators. A sensitivity analysis revealed that the segmentation step was the most influential image processing step on the final estimator error. Across all workflow scenarios, the RMSE of predicted nitrogen uptake ranged from 14.3 to 19.8 kg ha−1 (relative RMSE ranged from 26.5% to 36.5%), a 38.5% increase in error from the lowest to the highest error workflow scenario. The framework introduced demonstrates the sensitivity and extent to which image processing affects prediction accuracy. It allows remote sensing analysts to improve model performance while providing data-driven justification to improve the reproducibility and objectivity of their work, similar to the benefits of hyperparameter tuning in machine learning applications.

Modelling hyperspectral- and thermal-based plant traits for the early detection of Phytophthora-induced symptoms in oak decline

Holm oak decline is a complex phenomenon mainly influenced by the presence of Phytophthora cinnamomi and water stress. Plant functional traits (PTs) are altered during the decline process — initially affecting the physiological condition of the plants with non-visual symptoms and subsequently the leaf pigment content and canopy structure — being its quantification critical for the development of scalable detection methods for effective management. This study examines the relationship between spectral-based PTs and oak decline incidence and severity. We evaluate the use of high-resolution hyperspectral and thermal imagery (< 1 m) together with a 3-D radiative transfer model (RTM) to assess a supervised classification model of holm oak decline. Field surveys comprising more than 1100 trees with varying disease incidence and severity were used to train and validate the model and predictions. Declining trees showed decreases of model-based PTs such as water, chlorophyll, carotenoid, and anthocyanin contents, as well as fluorescence and leaf area index, and increases in crown temperature and dry matter content, compared to healthy trees. Our classification model built using different PT indicators showed up to 82% accuracy for decline detection and successfully identified 34% of declining trees that were not detected by visual inspection and confirmed in a re-evaluation 2 years later. Among all variables analysed, canopy temperature was identified as the most important variable in the model, followed by chlorophyll fluorescence. This methodological approach identified spectral plant traits suitable for the detection of pre-symptomatic trees and mapping of oak forest disease outbreaks up to 2 years in advance of identification via field surveys. Early detection can guide management activities such as tree culling and clearance to prevent the spread of dieback processes. Our study demonstrates the utility of 3-D RTM models to untangle the PT alterations produced by oak decline due to its heterogeneity. In particular, we show the combined use of RTM and machine learning classifiers to be an effective method for early detection of oak decline potentially applicable to many other forest diseases worldwide.

A FULLY AUTOMATED AND FAST APPROACH FOR CANOPY COVER ESTIMATION USING SUPER HIGH-RESOLUTION REMOTE SENSING IMAGERY

Abstract. Canopy cover is a key agronomic variable for understanding plant growth and crop development status. Estimation of canopy cover rapidly and accurately through a fully automated manner is significant with respect to high throughput plant phenotyping. In this work, we propose a simple, robust and fully automated approach, namely a rule-based method, that leverages the unique spectral pattern of green vegetation at visible (VIS) and near-infrared red (NIR) spectra regions to distinguish the green vegetation from background (i.e., soil, plant residue, non-photosynthetic vegetation leaves etc.), and then derive canopy cover. The proposed method was applied to high-resolution hyperspectral and multispectral imagery collected from gantry-based scanner and Unmanned Aerial Vehicle (UAV) platforms to estimate canopy cover. Additionally, machine learning methods, i.e., Support Vector Machine (SVM) and Random Forest (RF) were also employed as bench mark methods. The results show that: the rule-based method demonstrated promising classification accuracies that are comparable to SVM and RF for both hyperspectral and multispectral datasets. Although the rule-based method is more sensitive to mixed pixels and shaded canopy region, which potentially resulted in classification errors and underestimation of canopy cover in some cases; it showed better performance to detect smaller leaves than SVM and RF. Most importantly, the rule-based method substantially outperformed machine learning methods with respect to processing speed, indicating its greater potential for high-throughput plant phenotyping applications.

The effect of pixel heterogeneity for remote sensing based retrievals of evapotranspiration in a semi-arid tree-grass ecosystem

Many satellite missions rely on modeling approaches to acquire global or regional evapotranspiration (ET) products. However, a current challenge in ET modeling lies in dealing with sub-pixel heterogeneity, as models often assume homogeneous conditions at the pixel level. This is particularly an issue for heterogeneous landscapes, such as tree-grass ecosystems (TGE). In these areas, while appearing homogeneous at larger spatial scales pertaining to a single land cover type, the separation of the spectral signals of the main landscape features (e.g. trees and grasses) may not be achieved at the conventional satellite sensor resolution (e.g. 10–1000 m). This leads to important heterogeneity within the pixel grid that may not be accounted for in traditional modeling frameworks. This study examined the effect of pixel heterogeneity on ET simulations over a complex TGE in central Spain. High resolution hyperspectral imagery from five airborne campaigns forced the two-source energy balance (TSEB) model at 1.5–1000 m spatial resolutions. Along with this, the sharpened (20 m) and original (1000 m) Sentinels for Evapotranspiration (Sen-ET) products were evaluated over the study site for 2017. Results indicated that TSEB accurately simulated ET (RMSD: ~60 W/m2) when the pixel scale was able to robustly discriminate between grass and tree pixels (<5 m). However, model uncertainty drastically increased at spatial resolution greater than 10 m (RMSD: ~115 W/m2). Model performance remains relatively constant between 30 and 1000 m spatial resolutions, with within pixel heterogeneity being similar at all these scales. For mixed pixels (≥30 m), forcing an effective landscape roughness into TSEB (RMSD: ~80 W/m2) or applying a seasonally changing TSEB (TSEB-2S; RMSD: ~65 W/m2) improved the modeling performance. The Sen-ET products behaved similarly at both scales with RMSD of ET roughly 80 W/m2. The non-linear relationship between input parameters and flux output, along with the poor representation of aerodynamic surface roughness, were the main drivers for the increased uncertainties at coarser scales. These results suggest that care should be taken when using global ET products over TGE and similarly heterogeneous landscapes. The modeling procedure should inherently account for the presence of vastly different vegetation roughness elements within the pixel, to achieve reliable estimates of turbulent fluxes over a TGE.

Estimation of corn yield based on hyperspectral imagery and convolutional neural network

Corn is an important food crop in the world, widely distributed in many countries because of its excellent environmental adaptability. Moreover, corn is an important feed source for animal production and it is an indispensable raw material for many different industries. With increasing human population and decreasing arable land, there is an increased focus on increasing yield of corn.Convolutional neural network (CNN) analysis can be used for non-destructive yield prediction and is well suited for classification and feature extraction. The overall objective of this experiment was to use hyperspectral imagery to train a CNN classification model to estimate corn grain yield. High resolution hyperspectral imagery was captured at five corn growth stages - V5 (five leaves with visible leaf collars), V8 (eight leaves with visible leaf collars), V10 (ten leaves with visible leaf collars), V12 (12 leaves with visible leaf collars), and R2 (blister stage). Hyperspectral imagery was denoised using the wavelet analysis method, then was used to train and validate the CNN model. The spectral information reflecting the internal characteristics and the spatial information provided by the color image (red, green and blue bands extracted from hyperspectral image) reflecting the external characteristics of corn growth are extracted for modelling and verification. The results show that the spectral and color image-based integrated CNN model has a classification accuracy of 75.50%. In contrast, the accuracy of a one-dimensional CNN model based only on spectral information or a two-dimensional CNN model based only on color image information were 60.39% and 32.17%, respectively. The integrated CNN model (spectral information plus color image information) is better than results of the individual one-dimensional CNN or two-dimensional CNN models. In addition. The Kappa coefficient of integrated CNN model is 0.69, which indicates a high consistency of classification. Comprehensive use of spectral information and color image information, which represent information about the inner and outer corn canopy can provide more accurate corn yield prediction than one-dimensional or two-dimensional CNN models.

A Neural Network Method for Classification of Sunlit and Shaded Components of Wheat Canopies in the Field Using High-Resolution Hyperspectral Imagery

(1) Background: Information rich hyperspectral sensing, together with robust image analysis, is providing new research pathways in plant phenotyping. This combination facilitates the acquisition of spectral signatures of individual plant organs as well as providing detailed information about the physiological status of plants. Despite the advances in hyperspectral technology in field-based plant phenotyping, little is known about the characteristic spectral signatures of shaded and sunlit components in wheat canopies. Non-imaging hyperspectral sensors cannot provide spatial information; thus, they are not able to distinguish the spectral reflectance differences between canopy components. On the other hand, the rapid development of high-resolution imaging spectroscopy sensors opens new opportunities to investigate the reflectance spectra of individual plant organs which lead to the understanding of canopy biophysical and chemical characteristics. (2) Method: This study reports the development of a computer vision pipeline to analyze ground-acquired imaging spectrometry with high spatial and spectral resolutions for plant phenotyping. The work focuses on the critical steps in the image analysis pipeline from pre-processing to the classification of hyperspectral images. In this paper, two convolutional neural networks (CNN) are employed to automatically map wheat canopy components in shaded and sunlit regions and to determine their specific spectral signatures. The first method uses pixel vectors of the full spectral features as inputs to the CNN model and the second method integrates the dimension reduction technique known as linear discriminate analysis (LDA) along with the CNN to increase the feature discrimination and improves computational efficiency. (3) Results: The proposed technique alleviates the limitations and lack of separability inherent in existing pre-defined hyperspectral classification methods. It optimizes the use of hyperspectral imaging and ensures that the data provide information about the spectral characteristics of the targeted plant organs, rather than the background. We demonstrated that high-resolution hyperspectral imagery along with the proposed CNN model can be powerful tools for characterizing sunlit and shaded components of wheat canopies in the field. The presented method will provide significant advances in the determination and relevance of spectral properties of shaded and sunlit canopy components under natural light conditions.

Unsupervised Recurrent Hyperspectral Imagery Super-Resolution Using Pixel-Aware Refinement

Unsupervised fusion-based hyperspectral imagery (HSI) super-resolution (SR) is an essential task of HSI processing, which aims to reconstruct a high-resolution (HR) HSI using only an observed low-resolution HSI and a conventional HR image. Although a large number of unsupervised HSI SR methods have been proposed, the heuristic handcrafted image priors adopted by the majority of these methods restrict their capacity to capture specific characteristics of the HSI, as well as their ability to generalize to noisy observation images. In this study, we investigate a fusion-based HSI SR framework with the deep image prior, in which the deep neural network (rather than a heuristic handcrafted image prior) is exploited to capture plenty of image statistics. Within this framework, we further propose an unsupervised recurrence-based HSI SR method using pixel-aware refinement, which utilizes the intermediate reconstruction results to self-supervise unsupervised learning. Due to containing the information of the image-specific characteristic, the proposed method achieves better performance, in terms of both accuracy and robustness to noise, compared with the existing methods. Extensive experiments on four HSI data sets demonstrate the effectiveness of the proposed method.

Object based convolutional neural network for cloud classification in very high-resolution hyperspectral imagery

Remote sensing has a critical role for spatial data-based information systems and monitoring of the earth’s surface. The presence of clouds in optical sensing remote sensing satellite images is often a problem for many remote sensing applications. Therefore, the proper detection and classification of clouds in optical sensor remote sensing applications is quite a challenging task. This study aims to classify cloud objects in remote sensing satellite image data. The data used in this study is Pleiades very high-resolution satellite imagery data. The number of datasets used amounted to 1299 data. Cloud objects in this study are categorized into three classes, namely thick cloud, thin cloud, and clear. This study uses a deep learning algorithm, Convolutional Neural Network (CNN) for the classification of cloud objects. The CNN model used is LeNet with architectural modifications and parameters adjusted to the research needs. Classification of cloud objects with the LeNet model results in increased accuracy in each epoch during the training process and takes 1150.355 seconds for 200 epochs with the best accuracy value of 97.50%. The performance of LeNet is better than the VGG16 model as a comparison with the best accuracy of 96.50% with 600 data inputs.

Prediction of Early Season Nitrogen Uptake in Maize Using High-Resolution Aerial Hyperspectral Imagery

The ability to predict spatially explicit nitrogen uptake (NUP) in maize (Zea mays L.) during the early development stages provides clear value for making in-season nitrogen fertilizer applications that can improve NUP efficiency and reduce the risk of nitrogen loss to the environment. Aerial hyperspectral imaging is an attractive agronomic research tool for its ability to capture spectral data over relatively large areas, enabling its use for predicting NUP at the field scale. The overarching goal of this work was to use supervised learning regression algorithms—Lasso, support vector regression (SVR), random forest, and partial least squares regression (PLSR)—to predict early season (i.e., V6–V14) maize NUP at three experimental sites in Minnesota using high-resolution hyperspectral imagery. In addition to the spectral features offered by hyperspectral imaging, the 10th percentile Modified Chlorophyll Absorption Ratio Index Improved (MCARI2) was made available to the learning models as an auxiliary feature to assess its ability to improve NUP prediction accuracy. The trained models demonstrated robustness by maintaining satisfactory prediction accuracy across locations, pixel sizes, development stages, and a broad range of NUP values (4.8 to 182 kg ha−1). Using the four most informative spectral features in addition to the auxiliary feature, the mean absolute error (MAE) of Lasso, SVR, and PLSR models (9.4, 9.7, and 9.5 kg ha−1, respectively) was lower than that of random forest (11.2 kg ha−1). The relative MAE for the Lasso, SVR, PLSR, and random forest models was 16.5%, 17.0%, 16.6%, and 19.6%, respectively. The inclusion of the auxiliary feature not only improved overall prediction accuracy by 1.6 kg ha−1 (14%) across all models, but it also reduced the number of input features required to reach optimal performance. The variance of predicted NUP increased as the measured NUP increased (MAE of the Lasso model increased from 4.0 to 12.1 kg ha−1 for measured NUP less than 25 kg ha−1 and greater than 100 kg ha−1, respectively). The most influential spectral features were oftentimes adjacent to each other (i.e., within approximately 6 nm), indicating the importance of both spectral precision and derivative spectra around key wavelengths for explaining NUP. Finally, several challenges and opportunities are discussed regarding the use of these results in the context of improving nitrogen fertilizer management.

A Multiscale Deep Learning Approach for High-Resolution Hyperspectral Image Classification

Hyperspectral imagery (HSI) has emerged as a highly successful sensing modality for a variety of applications ranging from urban mapping to environmental monitoring and precision agriculture. Despite the efforts made by the scientific community, developing reliable algorithms of HSI classification remains a challenging problem, especially for high-resolution HSI data where there is often larger intraclass variability combined with a scarcity of the ground-truth data and class imbalance. In recent years, deep neural networks have emerged as a promising strategy for problems of HSI classification where they have shown remarkable potential for learning joint spectral-spatial features efficiently via backpropagation. In this letter, we propose a deep learning strategy for HSI classification that combines different convolutional neural networks, especially designed to efficiently learn joint spatial-spectral features over multiple scales. Our method achieves an overall classification accuracy of 66.73% on the 2018 IEEE GRSS hyperspectral data set-a high-resolution data set that includes 20 urban land-cover and land-use classes.

Mapping Succession in Non-Forest Habitats by Means of Remote Sensing: Is the Data Acquisition Time Critical for Species Discrimination?

The process of secondary succession is one of the most significant threats to non-forest (natural and semi-natural open) Natura 2000 habitats in Poland; shrub and tree encroachment taking place on abandoned, low productive agricultural areas, historically used as pastures or meadows, leads to changes to the composition of species and biodiversity loss, and results in landscape transformations. There is a perceived need to create a methodology for the monitoring of vegetation succession by airborne remote sensing, both from quantitative (area, volume) and qualitative (plant species) perspectives. This is likely to become a very important issue for the effective protection of natural and semi-natural habitats and to advance conservation planning. A key variable to be established when implementing a qualitative approach is the remote sensing data acquisition date, which determines the developmental stage of trees and shrubs forming the succession process. It is essential to choose the optimal date on which the spectral and geometrical characteristics of the species are as different from each other as possible. As part of the research presented here, we compare classifications based on remote sensing data acquired during three different parts of the growing season (spring, summer and autumn) for five study areas. The remote sensing data used include high-resolution hyperspectral imagery and LiDAR (Light Detection and Ranging) data acquired simultaneously from a common aerial platform. Classifications are done using the random forest algorithm, and the set of features to be classified is determined by a recursive feature elimination procedure. The results show that the time of remote sensing data acquisition influences the possibility of differentiating succession species. This was demonstrated by significant differences in the spatial extent of species, which ranged from 33.2% to 56.2% when comparing pairs of maps, and differences in classification accuracies, which when expressed in values of Cohen’s Kappa reached ~0.2. For most of the analysed species, the spring and autumn dates turned out to be slightly more favourable than the summer one. However, the final recommendation for the data acquisition time should take into consideration the phenological cycle of deciduous species present within the research area and the abiotic conditions.

Mapping hydrothermal alteration minerals using high-resolution AVIRIS-NG hyperspectral data in the Hutti-Maski gold deposit area, India

ABSTRACT The present study exploits high-resolution hyperspectral imagery acquired by the Airborne Visible/Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) sensor from the Hutti-Maski gold deposit area, India, to map hydrothermal alteration minerals. The study area is a volcanic-dominated late Archean greenstone belt that hosts major gold mineralization in the Eastern Dharwar Craton of southern India. The study encompasses pre-processing, spectral and spatial image reduction using Minimum Noise Fraction (MNF) and Fast Pixel Purity Index (FPPI), followed by endmember extraction using n-dimensional visualizer and the United States Geological Survey (USGS) mineral spectral library. Image derived endmembers such as goethite, chlorite, chlorite at the mine site (chlorite mixed with mined materials), kaolinite, and muscovite were subsequently used in spectral mapping methods such as Spectral Angle Mapper (SAM), Spectral Information Divergence (SID) and its hybrid, i.e. SIDSAMtan. Spectral similarity matrix of the target and non-target-based method has been proposed to find the possible optimum threshold needed to obtain mineral map using spectral mapping methods. Relative Spectral Discrimination Power (RSDPW) and Confusion Matrix (CM) have been used to evaluate the performance of SAM, SID, and SIDSAMtan. The RSDPW and CM illustrate that the SIDSAMtan benefits from the unique characteristics of SAM and SID to achieve better discrimination capability. The Overall Accuracy (OA) and kappa coefficient (ҡ) of SAM, SID, and SIDSAMtan were computed using 900 random validation points and obtained 90% (OA) and 0.88 (ҡ), 91.4% and 0.90, and 94.4% and 0.93, respectively. Obtained mineral map demonstrates that the northern portion of the area mainly consists of muscovite whereas the southern part is marked by chlorite, goethite, muscovite and kaolinite, indicating the propylitic alteration. Most of these minerals are associated with altered metavolcanic rocks and migmatite.

Retrieval of Salt Marsh Above-Ground Biomass from High-Spatial Resolution Hyperspectral Imagery Using PROSAIL

Salt marsh vegetation density varies considerably on short spatial scales, complicating attempts to evaluate plant characteristics using airborne remote sensing approaches. In this study, we used a mast-mounted hyperspectral imaging system to obtain cm-scale imagery of a salt marsh chronosequence on Hog Island, VA, where the morphology and biomass of the dominant plant species, Spartina alterniflora, varies widely. The high-resolution hyperspectral imagery allowed the detailed delineation of variations in above-ground biomass, which we retrieved from the imagery using the PROSAIL radiative transfer model. The retrieved biomass estimates correlated well with contemporaneously collected in situ biomass ground truth data ( R 2 = 0.73 ). In this study, we also rescaled our hyperspectral imagery and retrieved PROSAIL salt marsh biomass to determine the applicability of the method across spatial scales. Histograms of retrieved biomass changed considerably in characteristic marsh regions as the spatial scale of the imagery was progressively degraded. This rescaling revealed a loss of spatial detail and a shift in the mean retrieved biomass. This shift is indicative of the loss of accuracy that may occur when scaling up through a simple averaging approach that does not account for the detail found in the landscape at the natural scale of variation of the salt marsh system. This illustrated the importance of developing methodologies to appropriately scale results from very fine scale resolution up to the more coarse-scale resolutions commonly obtained in airborne and satellite remote sensing.

Using hyperspectral imaging to vector towards mineralization at the Canadian Malartic gold deposit, Québec, Canada

Canadian Malartic is a large-tonnage, low-grade Archean gold deposit (16.3 Moz, 1.08 g/t Au) located in the Abitibi region of Québec, Canada. A large part of the mineralization is hosted in the Pontiac Group metasedimentary rocks, which consist of mudstones to greywackes at upper greenschist to amphibolite facies. In exploration and production environments, these lithologies are challenging to characterize by conventional core logging methods, while in a research setting thin section-sized samples (2 cm × 4 cm) do not capture the full extent of mineralogical variability, which can extend from centimeters to meters away from mineralized zones. Here, high-resolution hyperspectral imagery (0.2–1.0 mm/pixel) in both shortwave infrared (SWIR, 1000–2500 nm) and longwave infrared (LWIR, 8000–12000 nm) is acquired for over two thousand meters of drill core, and is used to visualize changes in mineralogy and mineral chemistry related to metamorphism and hydrothermal alteration. Unaltered metasedimentary rocks contain metamorphic white mica with AlVI contents varying between 1.90 and 1.75 apfu (2195 to 2203 nm), depending on metamorphic grade. Hydrothermal alteration is characterized by white mica which becomes progressively more phengitic with increasing alteration intensity, with AlVI contents ranging from 1.70 to 1.50 apfu (2204 to 2212 nm). Phengitic white mica extends from meters to tens of meters away from major mineralized zones, and can be used as a vector towards mineralization in an exploration setting. White mica composition is correlated to Au content, and can be used to discriminate between unmineralized (<0.3 g/t Au, 2195–2204 nm), weakly mineralized (>0.3 g/t Au, 2205–2208 nm), and highly mineralized (>1.0 g/t Au, >2208 nm) samples, which is a simple metric that can directly be applied to sort ore in a production environment. The Mg# (molar Mg/[Mg + Fe]) of biotite, on the other hand, is unaffected by metamorphic grade, and consistently is Mg# 55–60 (2251–2250 nm) in unaltered samples. In mineralized samples, biotite is Mg-rich (Mg# >65, <2249 nm), but rapidly grades into background compositions outside of the ore zones. LWIR hyperspectral data is used in drill core to estimate silicification abundance from the strength of the 9200 nm quartz reflectance peak. Peak strength was generally in good agreement with the silicification abundance estimated by core logging, and is therefore a metric that can directly be used for characterization of this parameter in exploration and production environments. Over 1500 SWIR point measurements were also acquired in a 50 km by 15 km area around the deposit, and were used to delineate metamorphic isograds on regional scales, and a large 12 km by 3 km hydrothermal alteration halo around the Canadian Malartic deposit. Similar measurements, if performed along the over 200 km long Cadillac-Larder Lake Deformation zone, could be used to rapidly delineate other potentially mineralized areas.

Variational Pansharpening for Hyperspectral Imagery Constrained by Spectral Shape and Gram⁻Schmidt Transformation.

Image pansharpening can generate a high-resolution hyperspectral (HS) image by combining a high-resolution panchromatic image and a HS image. In this paper, we propose a variational pansharpening method for HS imagery constrained by spectral shape and Gram–Schmidt (GS) transformation. The main novelties of the proposed method are the additional spectral and correlation fidelity terms. First, we design the spectral fidelity term, which utilizes the spectral shape feature of the neighboring pixels with a new weight distribution strategy to reduce spectral distortion caused by the change in spatial resolution. Second, we consider that the correlation fidelity term uses the result of GS adaptive (GSA) to constrain the correlation, thereby preventing the low correlation between the pansharpened image and the reference image. Then, the pansharpening is formulized as the minimization of a new energy function, whose solution is the pansharpened image. In comparative trials, the proposed method outperforms GSA, guided filter principal component analysis, modulation transfer function, smoothing filter-based intensity modulation, the classic and the band-decoupled variational methods. Compared with the classic variation pansharpening, our method decreases the spectral angle from 3.9795 to 3.2789, decreases the root-mean-square error from 309.6987 to 228.6753, and also increases the correlation coefficient from 0.9040 to 0.9367.