Common Latent Space Exploration for Calibration Transfer across Hyperspectral Imaging-Based Phenotyping Systems

Hyperspectral imaging has increasingly been used in high-throughput plant phenotyping systems. Rapid advancement in the field of phenotyping has resulted in a wide array of hyperspectral imaging systems. However, sharing the plant feature prediction models between different phenotyping facilities becomes challenging due to the differences in imaging environments and imaging sensors. Calibration transfer between imaging facilities is crucially important to cope with such changes. Spectral space adjustment methods including direct standardization (DS), its variants (PDS, DPDS) and spectral scale transformation (SST) require the standard samples to be imaged in different facilities. However, in real-world scenarios, imaging the standard samples is practically unattractive. Therefore, in this study, we presented three methods (TCA, c-PCA, and di-PLSR) to transfer the calibration models without requiring the standard samples. In order to compare the performance of proposed approaches, maize plants were imaged in two greenhouse-based HTPP systems using two pushbroom-style hyperspectral cameras covering the visible near-infrared range. We tested the proposed methods to transfer nitrogen content (N) and relative water content (RWC) calibration models. The results showed that prediction R2 increased by up to 14.50% and 42.20%, while the reduction in RMSEv was up to 74.49% and 76.72% for RWC and N, respectively. The di-PLSR achieved the best results for almost all the datasets included in this study, with TCA being second. The performance of c-PCA was not at par with the di-PLSR and TCA. Our results showed that the di-PLSR helped to recover the performance of RWC, and N models plummeted due to the differences originating from new imaging systems (sensor type, spectrograph, lens system, spatial resolution, spectral resolution, field of view, bit-depth, frame rate, and exposure time) or lighting conditions. The proposed approaches can alleviate the requirement of developing a new calibration model for a new phenotyping facility or to resort to the spectral space adjustment using the standard samples.

INDIRECT INFORMATION HIDING TECHNOLOGY ON A MULTIADIC BASIS

It is shown that the current direction of increasing the safety of information resources when transmitting information in info-communication systems is the use of methods of steganographic instruction in video imagery. The effectiveness of such methods is significantly increased when used in a complex of methods of concealment, which are based on the principles of inconsistent and cosmic communication. At the same time, existing methods of steganographic are used in the process of insertion of information mainly only laws, empty features of visual perception of video images. So, it is justified that the scientific and applied problem, which is to increase the density of embedded messages in the video container with a given level of their reliability, is relevant. The solution of this problem is based on the solution of the contradiction, which concerns the fact that increasing the density of embedded data leads to a decrease in the bit rate of the video container, steganalysis stability, reliability of special information, and video container. Therefore, the research aims to develop a methodology for the steganographic embedding of information, taking into account the regularities of the video container, which are generated by its structural and structural-statistical features. The solution to the posed problem of applying steganographic transformations is proposed to be realised by methods of indirectly embedding parts of the hidden message in certain conditions or functional relationships. The possibility of creating steganographic transformations regarding the indirect embedding and extraction of hidden information in a multiadic basis by modifying the underlying basis system within an admissible set is demonstrated. It is shown that the multiadic system, which is created in the spectral space of DCT transforms, has the potential to form a set of admissible modifications of basis systems.

A Highly Accurate NILM: With an Electro-Spectral Space That Best Fits Algorithm’s National Deployment Requirements

The central problems of some of the existing Non-Intrusive Load Monitoring (NILM) algorithms are indicated as: (1) higher required electrical device identification accuracy; (2) the fact that they enable training over a larger device count; and (3) their ability to be trained faster, limiting them from usage in industrial premises and external grids due to their sensitivity to various device types found in residential premises. The algorithm accuracy is higher compared to previous work and is capable of training over at least thirteen electrical devices collaboratively, a number that could be much higher if such a dataset is generated. The algorithm trains the data around 1.8×108 faster due to a higher sampling rate. These improvements potentially enable the algorithm to be suitable for future “grids and industrial premises load identification” systems. The algorithm builds on new principles: an electro-spectral features preprocessor, a faster waveform sampling sensor, a shorter required duration for the recorded data set, and the use of current waveforms vs. energy load profile, as was the case in previous NILM algorithms. Since the algorithm is intended for operation in any industrial premises or grid location, fast training is required. Known classification algorithms are comparatively trained using the proposed preprocessor over residential datasets, and in addition, the algorithm is compared to five known low-sampling NILM rate algorithms. The proposed spectral algorithm achieved 98% accuracy in terms of device identification over two international datasets, which is higher than the usual success of NILM algorithms.

Investigating the Upper-Bound Performance of Sparse-Coding-Based Spectral Reconstruction from RGB Images

In Spectral Reconstruction (SR), we recover hyperspectral images from their RGB counterparts. Most of the recent approaches are based on Deep Neural Networks (DNN), where millions of parameters are trained mainly to extract and utilize the contextual features in large image patches as part of the SR process. On the other hand, the leading Sparse Coding method ‘A+’—which is among the strongest point-based baselines against the DNNs—seeks to divide the RGB space into neighborhoods, where locally a simple linear regression (comprised by roughly 102 parameters) suffices for SR. In this paper, we explore how the performance of Sparse Coding can be further advanced. We point out that in the original A+, the sparse dictionary used for neighborhood separations are optimized for the spectral data but used in the projected RGB space. In turn, we demonstrate that if the local linear mapping is trained for each spectral neighborhood instead of RGB neighborhood (and theoretically if we could recover each spectrum based on where it locates in the spectral space), the Sparse Coding algorithm can actually perform much better than the leading DNN method. In effect, our result defines one potential (and very appealing) upper-bound performance of point-based SR.

Coupling spectral and resource-use complementarity in experimental grassland and forest communities

Reflectance spectra provide integrative measures of plant phenotypes by capturing chemical, morphological, anatomical and architectural trait information. Here, we investigate the linkages between plant spectral variation, and spectral and resource-use complementarity that contribute to ecosystem productivity. In both a forest and prairie grassland diversity experiment, we delineated n -dimensional hypervolumes using wavelength bands of reflectance spectra to test the association between the spectral space occupied by individual plants and their growth, as well as between the spectral space occupied by plant communities and ecosystem productivity. We show that the spectral space occupied by individuals increased with their growth, and the spectral space occupied by plant communities increased with ecosystem productivity. Furthermore, ecosystem productivity was better explained by inter-individual spectral complementarity than by the large spectral space occupied by productive individuals. Our results indicate that spectral hypervolumes of plants can reflect ecological strategies that shape community composition and ecosystem function, and that spectral complementarity can reveal resource-use complementarity.

Potential and limits of vegetation indices compared to evaporite mineral indices for soil salinity discrimination and mapping

The study aims to analyze the ability of the most popular and widely used vegetation indices (VI’s), including NDVI, SAVI, EVI and TDVI, to discriminate and map soil salt contents compared to the potential of evaporite mineral indices such as SSSI and NDGI. The proposed methodology leverages on two complementary parts exploiting simulated and imagery data acquired over two study areas, i.e. Kuwait-State and Omongwa salt-pan in Namibia. In the first part, a field survey was conducted on the Kuwait site and 100 soil samples with various salinity levels and contents were collected; as well as, herbaceous vegetation cover canopy (alfalfa and forage plants) with various LAI coverage rates. In a Goniometric-Laboratory, the spectral signatures of all samples were measured and transformed using the continuum removed reflectance spectrum (CRRS) approach. Subsequently, they were resampled and convolved in the solar-reflective spectral bands of Landsat-OLI, and converted to the considered indices. Meanwhile, soil laboratory analyses were accomplished to measure pHs, electrical conductivity (EC-Lab), the major soluble cations and anions; thereby the sodium adsorption ratio was calculated. These elements support the investigation of the relationship between the spectral signature of each soil sample and its salt content. Furthermore, on the Omongwa salt-pan site, a Landsat-OLI image was acquired, pre-processed and converted to the investigated indices. Mineralogical ground-truth information collected during previous field work and an accurate Lidar DEM were used for the characterization and validation procedures on this second site. The obtained results demonstrated that regardless of the data source (simulation or image), the study site and the applied analysis methods, it is impossible for VI's to discriminate or to predict soil salinity. In fact, the spectral analysis revealed strong confusion between signals resulting from salt-crust and soil optical properties in the VNIR wavebands. The CRRS transformation highlighted the complete absence of salt absorption features in the blue, red and NIR wavelengths. As well as the analysis in 2D spectral-space pointed-out how VI’s compress and completely remove the signal fraction emitted by the soil background. Moreover, statistical regressions (p ˂ 0.05) between VI's and EC-Lab showed insignificant fits for SAVI, EVI and TDVI (R2 ≤ 0.06), and for NDVI (R2 of 0.35). Although the Omongwa is a natural flat salt playa, the four derived VI’s from OLI image are completely unable to detect the slightest grain of salt in the soil. Contrariwise, analyses of spectral signatures and CRRS highlighted the potential of the SWIR spectral domain to distinguish salt content in soil regardless of its optical properties. Likewise, according to Kuwait spectral data and EC-Lab analysis, NDGI and SSSI incorporating SWIR wavebands have performed very well and similarly (R2 of 0.72) for the differentiation of salt-affected soil classes. These statistical results were also corroborated visually by the maps derived from these evaporite indices over the salt-pan site, as well as by their consistency with the validation points representing the ground truth. However, although both the indices have mapped the salinity patterns almost similarly, NDGI further highlights the gypsum content. While the SSSI show greater sensitivity to salt crusts present in the pan area that are formed from different mineral sources (i.e., halite, gypsum, etc.).

Validation of Multi-Frame PIV Image Interrogation Algorithms in the Spectral Domain

Multi-frame correlation algorithms for time-resolved PIV have been shown in previous studies to reduce noise and error levels in comparison with conventional two-frame correlations. However, none of these prior efforts tested the accuracy of the algorithms in spectral space. Even should a multi-frame algorithm reduce the error of vector computations summed over an entire data set, this does not imply that these improvements are observed at all frequencies. The present study examines the accuracy of velocity spectra in comparison with simultaneous hot-wire data. Results indicate that the high-frequency content of the spectrum is very sensitive to choice of the interrogation algorithm and may not return an accurate response. A top-hat-weighted sliding sum-of-correlation is contaminated by high-frequency ringing whereas Gaussian weighting is indistinguishable from a low-pass filtering effect. Some evidence suggests the pyramid correlation modestly increases bandwidth of the measurement at high frequencies. The apparent benefits of multi-frame interrogation algorithms may be limited in their ability to reveal additional spectral content of the flow.

Reduced Multiplicative (BURA-MR) and Additive (BURA-AR) Best Uniform Rational Approximation Methods and Algorithms for Fractional Elliptic Equations

Numerical methods for spectral space-fractional elliptic equations are studied. The boundary value problem is defined in a bounded domain of general geometry, Ω⊂Rd, d∈{1,2,3}. Assuming that the finite difference method (FDM) or the finite element method (FEM) is applied for discretization in space, the approximate solution is described by the system of linear algebraic equations Aαu=f, α∈(0,1). Although matrix A∈RN×N is sparse, symmetric and positive definite (SPD), matrix Aα is dense. The recent achievements in the field are determined by methods that reduce the original non-local problem to solving k auxiliary linear systems with sparse SPD matrices that can be expressed as positive diagonal perturbations of A. The present study is in the spirit of the BURA method, based on the best uniform rational approximation rα,k(t) of degree k of tα in the interval [0,1]. The introduced additive BURA-AR and multiplicative BURA-MR methods follow the observation that the matrices of part of the auxiliary systems possess very different properties. As a result, solution methods with substantially improved computational complexity are developed. In this paper, we present new theoretical characterizations of the BURA parameters, which gives a theoretical justification for the new methods. The theoretical estimates are supported by a set of representative numerical tests. The new theoretical and experimental results raise the question of whether the almost optimal estimate of the computational complexity of the BURA method in the form O(Nlog2N) can be improved.