Hyperspectral Acquisitions Research Articles

Unmanned aerial vehicle and proximal sensing of vegetation indices in olive tree (<i>Olea europaea</i>)

Remote and proximal sensing platforms at the service of precision olive growing are bringing new development possibilities to the sector. A proximal sensing platform is close to the vegetation, while a remote sensing platform, such as unmanned aerial vehicle (UAV), is more distant but has the advantage of rapidity to investigate plots. The study aims to compare multispectral and hyperspectral data acquired with remote and proximal sensing platforms. The comparison between the two sensors aims at understanding the different responses their use can provide on a crop, such as olive trees having a complex canopy. The multispectral data were acquired with a DJI multispectral camera mounted on the UAV Phantom 4. Hyperspectral acquisitions were carried out with a FieldSpec® HandHeld 2™ Spectroradiometer in the canopy portions exposed to South, East, West, and North. The multispectral images were processed with Geographic Information System software to extrapolate spectral information for each cardinal direction’s exposure. The three main Vegetation indices were used: normalized difference vegetation index (NDVI), normalized difference red-edge index (NDRE), and modified soil adjusted vegetation index (MSAVI). Multispectral data could describe the total variability of the whole plot differentiating each single plant status. Hyperspectral data were able to describe vegetation conditions more accurately; they appeared to be related to the cardinal exposure. MSAVI, NDVI, and NDRE showed correlation r =0.63**, 0.69**, and 0.74**, respectively, between multispectral and hyperspectral data. South and West exposures showed the best correlations with both platforms.

A Novel Spectral Index to Identify Cacti in the Sonoran Desert at Multiple Scales Using Multi-Sensor Hyperspectral Data Acquisitions

Accurate identification of cacti, whether seen as an indicator of ecosystem health or an invasive menace, is important. Technological improvements in hyperspectral remote sensing systems with high spatial resolutions make it possible to now monitor cacti around the world. Cacti produce a unique spectral signature because of their morphological and anatomical characteristics. We demonstrate in this paper that we can leverage a reflectance dip around 972 nm, due to cacti’s morphological structure, to distinguish cacti vegetation from non-cacti vegetation in a desert landscape. We also show the ability to calculate two normalized vegetation indices that highlight cacti. Furthermore, we explore the impacts of spatial resolution by presenting spectral signatures from cacti samples taken with a handheld field spectroradiometer, drone-based hyperspectral sensor, and aerial hyperspectral sensor. These cacti indices will help measure baseline levels of cacti around the world and examine changes due to climate, disturbance, and management influences.

Multitemporal hyperspectral tree species classification in the Białowieża Forest World Heritage site

Abstract Tree species composition maps derived from hyperspectral data have been found to be accurate but it is still unclear whether an optimal time window exists to acquire the images. Trees in temperate forests are subject to phenological changes that are species-specific and can have an impact on species recognition. Our study examined the performance of a multitemporal hyperspectral dataset to classify tree species in the Polish part of the Białowieża Forest. We classified seven tree species including spruce (Picea abies (L.) H.Karst), pine (Pinus sylvestris L.), alder (Alnus glutinosa Gaertn.), oak (Quercus robur L.), birch (Betula pendula Roth), hornbeam (Carpinus betulus L.) and linden (Tilia cordata Mill.), using Support Vector Machines. We compared the results for three data acquisitions—early and late summer (2–4 July and 24–27 August), and autumn (1–2 October) as well as a classification based on an image stack containing all three acquisitions. Furthermore, the sizes (height and crown diameter) of misclassified and correctly classified trees of the same species were compared. The early summer acquisition reached the highest accuracies with an Overall Accuracy (OA) of 83–94 per cent and Kappa (κ) of 0.80–0.92. The classification based on the stacked multitemporal dataset resulted in slightly higher accuracies (84–94 per cent OA and 0.81–0.92 κ). For some species, e.g. birch and oak, tree size differed notably for correctly and incorrectly classified trees. We conclude that implementing multitemporal hyperspectral data can improve the classification result as compared with a single acquisition. However, the obtained accuracy of the multitemporal image stack was in our case comparable to the best single-date classification and investing more time in identifying regionally optimal acquisition windows may be worthwhile as long hyperspectral acquisitions are still sparse.

DETECTION AND QUANTIFICATION OF INDUSTRIAL METHANE PLUME WITH THE AIRBORNE HYSPEX-NEO CAMERA AND APPLICATIONS TO SATELLITE DATA

Abstract. The control and monitoring of greenhouse gases is an important issue for the study of climate change, for the impact in terms of public health or for the risks related to industry. An algorithm has been developped dedicated to commercial airborne hyperspectral camera as HySpex-NEO for the detection of industrial methane plume and the quantification of the emission source. HySpex-NEO is an imager used in airborne campaign with a spatial resolution of 1.4 m at flight altitude of 2 km, a swath of 650 m and a spectral resolution of 6 nm. This algorithm has been validated over a controlled release less than 100 g/s during an airborne campaign over Lacq (France) industry. It has also been applied to the Aliso Canyon leakage data acquired with AVIRIS JPL (Airborne Visible Infrared Spectrometer) with a spatial resolution of 6.8 m at flight altitude of 6 km, a swath of 5.6 km and a spectral resolution of 10 nm. Application to satellite hyperspectral data is shown on artificial data derived from airborne hyperspectral acquisitions.

Retrieval of aboveground crop nitrogen content with a hybrid machine learning method

Hyperspectral acquisitions have proven to be the most informative Earth observation data source for the estimation of nitrogen (N) content, which is the main limiting nutrient for plant growth and thus agricultural production. In the past, empirical algorithms have been widely employed to retrieve information on this biochemical plant component from canopy reflectance. However, these approaches do not seek for a cause-effect relationship based on physical laws. Moreover, most studies solely relied on the correlation of chlorophyll content with nitrogen, and thus neglected the fact that most N is bound in proteins. Our study presents a hybrid retrieval method using a physically-based approach combined with machine learning regression to estimate crop N content. Within the workflow, the leaf optical properties model PROSPECT-PRO including the newly calibrated specific absorption coefficients (SAC) of proteins, was coupled with the canopy reflectance model 4SAIL to PROSAIL-PRO. The latter was then employed to generate a training database to be used for advanced probabilistic machine learning methods: a standard homoscedastic Gaussian process (GP) and a heteroscedastic GP regression that accounts for signal-to-noise relations. Both GP models have the property of providing confidence intervals for the estimates, which sets them apart from other machine learners. Moreover, a GP-based sequential backward band removal algorithm was employed to analyze the band-specific information content of PROSAIL-PRO simulated spectra for the estimation of aboveground N. Data from multiple hyperspectral field campaigns, carried out in the framework of the future satellite mission Environmental Mapping and Analysis Program (EnMAP), were exploited for validation. In these campaigns, corn and winter wheat spectra were acquired to simulate spectral EnMAP data. Moreover, destructive N measurements from leaves, stalks and fruits were collected separately to enable plant-organ-specific validation. The results showed that both GP models can provide accurate aboveground N simulations, with slightly better results of the heteroscedastic GP in terms of model testing and against in situ N measurements from leaves plus stalks, with root mean square error (RMSE) of 2.1 g/m². However, the inclusion of fruit N content for validation deteriorated the results, which can be explained by the inability of the radiation to penetrate the thick tissues of stalks, corn cobs and wheat ears. GP-based band analysis identified optimal spectral settings with ten bands mainly situated in the shortwave infrared (SWIR) spectral region. Use of well-known protein absorption bands from the literature showed comparative results. Finally, the heteroscedastic GP model was successfully applied on airborne hyperspectral data for N mapping. We conclude that GP algorithms, and in particular the heteroscedastic GP, should be implemented for global agricultural monitoring of aboveground N from future imaging spectroscopy data.

Combining the use of CNN classification and strength‐driven compression for the robust identification of bacterial species on hyperspectral culture plate images

Huge streams of diagnostic images are expected to be produced daily in the emerging field of digital microbiology imaging because of the ongoing worldwide spread of Full Laboratory Automation systems. This is redefining the way microbiologists execute diagnostic tasks. In this context, the authors want to assess the suitability and effectiveness of a deep learning approach to solve the diagnostically relevant but visually challenging task of directly identifying pathogens on bacterial growing plates. In particular, starting from hyperspectral acquisitions in the VNIR range and spatial‐spectral processing of cultured plates, they approach the identification problem as the classification of computed spectral signatures of the bacterial colonies. In a highly relevant clinical context (urinary tract infections) and on a database of acquired hyperspectral images, they designed and trained a convolutional neural network for pathogen identification, assessing its performance and comparing it against conventional classification solutions. At the same time, given the expected data flow and possible conservation and transmission needs, they are interested in evaluating the combined use of classification and lossy data compression. To this end, after selecting a suitable wavelet‐based compression technology, they test coding strength‐driven operating points looking for configurations able to provably prevent any classification performance degradation.

On the Effect of Spatially Non-Disjoint Training and Test Samples on Estimated Model Generalization Capabilities in Supervised Classification With Spatial Features

In this letter, we establish two sampling schemes to select training and test sets for supervised classification. We do this in order to investigate whether estimated generalization capabilities of learned models can be positively biased from the use of spatial features. Numerous spatial features impose homogeneity constraints on the image data, whereby a spatially connected set of image elements is attributed identical feature values. In addition to a frequent occurrence of intrinsic spatial autocorrelation, this leads to extrinsic spatial autocorrelation with respect to the image data. The first sampling scheme follows a spatially random partitioning into training and test sets. In contrast to that, the second strategy implements a spatially disjoint partitioning, which considers in particular topological constraints that arise from the deployment of spatial features. Experimental results are obtained from multi- and hyperspectral acquisitions over urban environments. They underline that a large share of the differences between estimated generalization capabilities obtained with the spatially disjoint and non-disjoint sampling strategies can be attributed to the use of spatial features, whereby differences increase with an increasing size of the spatial neighborhood considered for computing a spatial feature. This stresses the necessity of a proper spatial sampling scheme for model evaluation to avoid overoptimistic model assessments.

A regression approach to the mapping of bio-physical characteristics of surface sediment using in situ and airborne hyperspectral acquisitions

Remote sensing has been successfully utilized to distinguish and quantify sediment properties in the intertidal environment. Classification approaches of imagery are popular and powerful yet can lead to site- and case-specific results. Such specificity creates challenges for temporal studies. Thus, this paper investigates the use of regression models to quantify sediment properties instead of classifying them. Two regression approaches, namely multiple regression (MR) and support vector regression (SVR), are used in this study for the retrieval of bio-physical variables of intertidal surface sediment of the IJzermonding, a Belgian nature reserve. In the regression analysis, mud content, chlorophyll a concentration, organic matter content, and soil moisture are estimated using radiometric variables of two airborne sensors, namely airborne hyperspectral sensor (AHS) and airborne prism experiment (APEX) and and using field hyperspectral acquisitions by analytical spectral device (ASD). The performance of the two regression approaches is best for the estimation of moisture content. SVR attains the highest accuracy without feature reduction while MR achieves good results when feature reduction is carried out. Sediment property maps are successfully obtained using the models and hyperspectral imagery where SVR used with all bands achieves the best performance. The study also involves the extraction of weights identifying the contribution of each band of the images in the quantification of each sediment property when MR and principal component analysis are used.

Semisupervised Transfer Component Analysis for Domain Adaptation in Remote Sensing Image Classification

In this paper, we study the problem of feature extraction for knowledge transfer between multiple remotely sensed images in the context of land-cover classification. Several factors such as illumination, atmospheric, and ground conditions cause radiometric differences between images of similar scenes acquired on different geographical areas or over the same scene but at different time instants. Accordingly, a change in the probability distributions of the classes is observed. The purpose of this work is to statistically align in the feature space an image of interest that still has to be classified (the target image) to another image whose ground truth is already available (the source image). Following a specifically designed feature extraction step applied to both images, we show that classifiers trained on the source image can successfully predict the classes of the target image despite the shift that has occurred. In this context, we analyze a recently proposed domain adaptation method aiming at reducing the distance between domains, Transfer Component Analysis, and assess the potential of its unsupervised and semisupervised implementations. In particular, with a dedicated study of its key additional objectives, namely the alignment of the projection with the labels and the preservation of the local data structures, we demonstrate the advantages of Semisupervised Transfer Component Analysis. We compare this approach with other both linear and kernel-based feature extraction techniques. Experiments on multi- and hyperspectral acquisitions show remarkable cross- image classification performances for the considered strategy, thus confirming its suitability when applied to remotely sensed images.