RESUMEN

Since water is an essential resource in various fields, it requires constant monitoring. Chlorophyll-a concentration is a crucial indicator of water quality and can be used to monitor water quality. In this study, we developed methods to forecast chlorophyll-a concentrations in real-time using hyperspectral data on IoT platform and various machine learning algorithms. Compared to regular cameras that record information only in the three broad color bands of red, green, and blue, the hyperspectral images of drinking water sources record the data in dozens or even hundreds of distinct small wavelength bands, providing each pixel in an image with a full spectrum. Different machine learning algorithms have been developed using hyperspectral data and field observations of water quality and weather conditions. Previous studies have predicted chlorophyll concentrations using either partial least squares (PLS), which is a dimensionality reduction method, or machine learning. In contrast, our study employed the PLS technique as a preprocessing step to diminish the dimensionality of the hyperspectral data, followed by the application of the machine learning techniques with optimized hyperparameters to improve the precision of the predictions, thereby introducing a real-time mechanism for chlorophyll-a prediction. Consequently, a machine learning algorithm with R2 values of 0.9 or above and sufficiently small RMSE was developed for real-time chlorophyll-a forecasting. Real-time chlorophyll-a forecasting using LightGBM has the best performance, with a mean R2 of 0.963 and a mean RMSE of 2.679. This paper is expected to have applications in algal bloom early detection on monitoring systems.

RESUMEN

Aboveground biomass (AGB) is regarded as a critical variable in monitoring crop growth and yield. The use of hyperspectral remote sensing has emerged as a viable method for the rapid and precise monitoring of AGB. Due to the extensive dimensionality and volume of hyperspectral data, it is crucial to effectively reduce data dimensionality and select sensitive spectral features to enhance the accuracy of rice AGB estimation models. At present, derivative transform and feature selection algorithms have become important means to solve this problem. However, few studies have systematically evaluated the impact of derivative spectrum combined with feature selection algorithm on rice AGB estimation. To this end, at the Xiaogang Village (Chuzhou City, China) Experimental Base in 2020, this study used an ASD FieldSpec handheld 2 ground spectrometer (Analytical Spectroscopy Devices, Boulder, Colorado, USA) to obtain canopy spectral data at the critical growth stage (tillering, jointing, booting, heading, and maturity stages) of rice, and evaluated the performance of the recursive feature elimination (RFE) and Boruta feature selection algorithm through partial least squares regression (PLSR), principal component regression (PCR), support vector machine (SVM) and ridge regression (RR). Moreover, we analyzed the importance of the optimal derivative spectrum. The findings indicate that (1) as the growth stage progresses, the correlation between rice canopy spectrum and AGB shows a trend from high to low, among which the first derivative spectrum (FD) has the strongest correlation with AGB. (2) The number of feature bands selected by the Boruta algorithm is 19~35, which has a good dimensionality reduction effect. (3) The combination of FD-Boruta-PCR (FB-PCR) demonstrated the best performance in estimating rice AGB, with an increase in R² of approximately 10% ~ 20% and a decrease in RMSE of approximately 0.08% ~ 14%. (4) The best estimation stage is the booting stage, with R2 values between 0.60 and 0.74 and RMSE values between 1288.23 and 1554.82 kg/hm2. This study confirms the accuracy of hyperspectral remote sensing in estimating vegetation biomass and further explores the theoretical foundation and future direction for monitoring rice growth dynamics.

RESUMEN

Wetlands cover only around 6 % of the Earth's land surface, and are recognized as one of the three major ecosystems, alongside forests and oceans. The ecological structure and function of karst wetlands are unique due to the influence of geologic structure. At present, the unclear spectral morphology of surface water in karst wetlands poses a significant challenge in remote sensing estimation of non-optically active water quality parameters (NAWQPs). This study proposed a novel multi-scale spectral morphology feature extraction (MSFE) method to insight to spectral characteristics in surface water of karst wetlands, and further screen the sensitive features of NAWQPs. Then we constructed three remote sensing inversion strategies for NAWQPs (TN, TP, NH3_N, DO), including direct estimation, indirect estimation, and auxiliary estimation. Finally, we constructed a novel pH-based hierarchical analysis framework (pH_HA) to thoroughly explore the influence of alkalinity-biased characteristics of karst water on the spectral domain of NAWQPs and its estimation accuracy using in-situ hyperspectral data, respectively. We found that the spectral characteristics of karst waters at the first reflectance peak (580 nm) differed significantly from other water body types. The MSFE successfully captured the sensitive spectral domains for NAWQPs, and focused on between 500 and 600 nm and 900-960 nm. The sensitive features captured by MSFE improved estimation accuracy of NAWQPs (R2 >0.9). Direct estimation presented more stable performance compared to the auxiliary estimation (average RMSE of 0.366 mg/L), and the auxiliary estimation model further improved the retrieval accuracy of TN compared to direct estimation model (R2 increasing from 0.43 to 0.56). The novel hierarchical framework clearly revealed the notable changes in the sensitive spectral domains of NAWQPs under different pH values, and enabled more precise determination of spectral subdomains of NAWQPs, and identified the optimal spectral features. The pH_HA framework effectively improved the estimation accuracy of NAWQPs (R2 increased from 0.514 to over 0.9), and the estimation accuracies (R2) of four NAWQPs were all more than 0.9 when the pH value was over 8.5. Our works provide an effective approach for monitoring water quality in karst wetlands.

Asunto(s)

RESUMEN

The diversity of fungal strains is influenced by genetic and environmental factors, growth conditions and mycelium age, and the spectral features of fungal mycelia are associated with their biochemical, physiological, and structural traits. This study investigates whether intraspecific differences can be detected in two closely related entomopathogenic species, namely Cordyceps farinosa and Cordyceps fumosorosea, using ultraviolet A to shortwave infrared (UVA-SWIR) reflectance spectra. Phylogenetic analysis of all strains revealed a high degree of uniformity among the populations of both species. The characteristics resulting from variation in the species, as well as those resulting from the age of the cultures were determined. We cultured fungi on PDA medium and measured the reflectance of mycelia in the 350-2500 nm range after 10 and 17 days. We subjected the measurements to quadratic discriminant analysis (QDA) to identify the minimum number of bands containing meaningful information. We found that when the age of the fungal culture was known, species represented by a group of different strains could be distinguished with no more than 3-4 wavelengths, compared to 7-8 wavelengths when the age of the culture was unknown. At least 6-8 bands were required to distinguish cultures of a known species among different age groups. Distinguishing all strains within a species was more demanding: at least 10 bands were required for C. fumosorosea and 21 bands for C. farinosa. In conclusion, fungal differentiation using point reflectance spectroscopy gives reliable results when intraspecific and age variations are taken into account.

Asunto(s)

RESUMEN

Fourier-transform infrared spectroscopy (FTIR) is a powerful, non-destructive, highly sensitive and a promising analytical technique to provide spectrochemical signatures of biological samples, where markers like carbohydrates, proteins, and phosphate groups of DNA can be recognized in biological micro-environment. However, method of measurements of large cells need an excessive time to achieve high quality images, making its clinical use difficult due to speed of data-acquisition and lack of optimized computational procedures. To address such challenges, Machine Learning (ML) based technologies can assist to assess an accurate prognostication of breast cancer (BC) subtypes with high performance. Here, we applied FTIR spectroscopy to identify breast cancer subtypes in order to differentiate between luminal (BT474) and non-luminal (SKBR3) molecular subtypes. For this reason, we tested multivariate classification technique to extract feature information employing three-dimension (3D)-discriminant analysis approach based on 3D-principle component analysis-linear discriminant analysis (3D-PCA-LDA) and 3D-principal component analysis-quadratic discriminant analysis (3D-PCA-QDA), showing an improvement in sensitivity (98%), specificity (94%) and accuracy (98%) parameters compared to conventional unfolded methods. Our results evidence that 3D-PCA-LDA and 3D-PCA-QDA are potential tools for discriminant analysis of hyperspectral dataset to obtain superior classification assessment.

Asunto(s)

RESUMEN

The leaf chlorophyll content (LCC) of vegetation is closely related to photosynthetic efficiency and biological activity. Jujube (Ziziphus jujuba Mill.) is a traditional economic forest tree species. Non-destructive monitoring of LCC of jujube is of great significance for guiding agroforestry production and promoting ecological environment protection in arid and semi-arid lands. Hyperspectral data is an important data source for LCC detection. However, hyperspectral data consists of a multitude of bands and contains extensive information. As a result, certain bands may exhibit high correlation, leading to redundant spectral information. This redundancy can distort LCC prediction results and reduce accuracy. Therefore, it is crucial to select appropriate preprocessing methods and employ effective data mining techniques when analyzing hyperspectral data. This study aims to evaluate the performance of hyperspectral data for estimating LCC of jujube trees by integrating different derivative processing techniques with different dimensionality reduction algorithms. Hyperspectral reflectance data were obtained through simulations using an invertible forest reflectance model (INFORM) and measurements from jujube tree canopies. The least absolute shrinkage and selection operator (LASSO) and elastic net (EN) were employed to identify the important bands in the original spectra (OS), first derivative spectra (FD), and second derivative spectra (SD). Support vector regression (SVR) was used to establish the estimation model. The results show that compared with full-spectrum modeling, LASSO and EN algorithms are effective methods for preventing overfitting in LCC machine learning estimation models for different spectral derivatives. The LASSO/EN-based estimation models constructed using FD and SD exhibited superior R2 compared to the OS. The important band of SD can best reveal the relevant information of jujube LCC, and SD-EN-SVR is the most ideal model in both the simulated dataset (R2 = 0.99, RMSE=0.61) and measured dataset (R2 = 0.89, RMSE=0.91). Our results provided a reference for rapid and non-destructive estimation of the LCC of agroforestry vegetation using canopy hyperspectral data.

RESUMEN

Soil salinization is one of the main causes of land degradation in arid and semi-arid areas. Timely and accurate monitoring of soil salinity in different areas is a prerequisite for amelioration. Hyperspectral technology has been widely used in soil salinity monitoring due to its high efficiency and rapidity. However, vegetation cover is an inevitable interference in the direct acquisition of soil spectra during crop growth period, which greatly limits the monitoring of soil salinity by remote sensing. Due to high soil salinity could lead to difficulty in plants' water absorption, and inhibit plant dry matter accumulation, a method for monitoring root zone soil salinity by combining vegetation canopy spectral information and crop aboveground growth parameters was proposed in this study. The canopy spectral information was acquired by a spectroradiometer, and then variable importance in projection (VIP), competitive adaptive reweighted sampling (CARS), and random frog algorithm (RFA) were used to extract the salinity spectral features in cotton canopy spectrum. The extracted features were then used to estimate root zone soil salinity in cotton field by combining with cotton plant height, aboveground biomass, and shoot water content. The results showed that there was a negative correlation between plant height/aboveground biomass/shoot water content and soil salinity in 0-20, 0-40, and 0-60 cm soil layers at different growth stages of cotton. Spectral feature selection by the three methods all improved the prediction accuracy of soil salinity, especially CARS. The prediction accuracy based on the combination of spectral features and cotton growth parameters was significantly higher than that based on only spectral features, with R2 increasing by 10.01%, 18.35%, and 29.90% for the 0-20, 0-40, and 0-60 cm soil layer, respectively. The model constructed based on the first derivative spectral preprocessing, spectral feature selection by CARS, cotton plant height, and shoot water content had the highest accuracy for each soil layer, with R2 of 0.715,0.769, and 0.742 for the 0-20, 0-40, 0-60 cm soil layer, respectively. Therefore, the method by combining cotton canopy hyperspectral data and plant growth parameters could significantly improve the prediction accuracy of root zone soil salinity under vegetation cover conditions. This is of great significance for the amelioration of saline soil in salinized farmlands arid areas.

RESUMEN

Soluble solids content is an important evaluation index affecting the quality of greengage fruit. The SSC content of green plum determines the picking time of green plum and what products are finally made into the market, such as preserves or fruit wine. The traditional destructive experiment is not conducive to the subsequent processing of green plum, and the efficiency is low and the labor cost is high. In this paper, hyperspectral images of green plums are analyzed based on the DenseNet network model, and a sugar content prediction model for green plums is established. After experimental collection and screening, 366 samples were obtained for the prediction of sugar content. According to the ratio of 3:1, 274 samples were obtained for the training set and 92 samples for the test set. In the prediction of sugar content, compared with the PLSR and MobileNetV2 model, the Rp of the 1D-DenseNet121 model in this experiment increased by 8.95%, and 6.27% respectively. and the MAEp was reduced by 15.44% and 10.35% respectively. The 1D-DenseNet121 model had a faster iterative convergence rate than the MobileNetV2 model, showing better prediction performance, which is more in line with the actual demand for green plum sorting, effectively improving the low efficiency of traditional physical and chemical detection.

RESUMEN

The continuous availability of spatial and temporal distributed data from satellite sensors provides more accurate and timely information regarding surface water quality parameters. Remote sensing data has the potential to serve as an alternative to traditional on-site measurements, which can be resource-intensive due to the time and labor involved. This present study aims in exploring the possibility and comparison of hyperspectral and multispectral imageries (PRISMA) for accurate prediction of surface water quality parameters. Muthupet estuary, situated on the south side of the Cauvery River delta on the Bay of Bengal, is selected as the study area. The remote sensing data is acquired from the PRISMA hyperspectral satellite and the Sentinel-2 multispectral instrument (MSI) satellite. The in situ sampling from the study area is performed, and the testing procedures are carried out for analyzing different water quality parameters. The correlations between the water sample results and the reflectance values of satellites are analyzed to generate appropriate algorithmic models. The study utilized data from both the PRISMA and Sentinel satellites to develop models for assessing water quality parameters such as total dissolved solids, chlorophyll, pH, and chlorides. The developed models demonstrated strong correlations with R2 values above 0.80 in the validation phase. PRISMA-based models for pH and chlorophyll displayed higher accuracy levels than Sentinel-based models with R2 > 0.90.

Asunto(s)

RESUMEN

In recent years, different deep learning frameworks were introduced for hyperspectral image (HSI) classification. However, the proposed network models have a higher model complexity, and do not provide high classification accuracy if few-shot learning is used. This paper presents an HSI classification method that combines random patches network (RPNet) and recursive filtering (RF) to obtain informative deep features. The proposed method first convolves image bands with random patches to extract multi-level deep RPNet features. Thereafter, the RPNet feature set is subjected to dimension reduction through principal component analysis (PCA), and the extracted components are filtered using the RF procedure. Finally, the HSI spectral features and the obtained RPNet-RF features are combined to classify the HSI using a support vector machine (SVM) classifier. In order to test the performance of the proposed RPNet-RF method, some experiments were performed on three widely known datasets using a few training samples for each class, and classification results were compared with those obtained by other advanced HSI classification methods adopted for small training samples. The comparison showed that the RPNet-RF classification is characterized by higher values of such evaluation metrics as overall accuracy and Kappa coefficient.

RESUMEN

Overlapping fingermark images are sometimes discarded because fingermark collation for the individual fingermarks is difficult. Fluorescence hyperspectral data (HSD) measured using the models of double overlapping fingermarks obtained under the excitation of a high-power, continuous wave, green laser is suitable for obtaining individual fingermark images. However, there are limitations such as the problems on each spectrum of the individual fingermark and the forensic value of the obtained images. In this study, independent component analysis (ICA) was applied to the fluorescence HSD obtained from the models of doubly overlapping fingermarks, to obtain independent component (IC) spectra and the corresponding IC images. Forensic value of the obtained IC images was examined, considering the possibility of fingermark collation in masked fashion to the model sample information. The IC images obtained from the HSD had enough potential to enable extracting twelve minutiae required for fingermark collation if the image quality was good.

RESUMEN

PROBLEM: Similarity measures are widely used as an approved method for spectral discrimination or identification with their applications in different areas of scientific research. Even though a range of works have been presented, only a few showed slightly promising results for human tissue, and these were mostly focused on pathological and non-pathological tissue classification. METHODS: In this work, several spectral similarity measures on hyperspectral (HS) images of in vivo human tissue were evaluated for tissue discrimination purposes. Moreover, we introduced two new hybrid spectral measures, called SID-JM-TAN(SAM) and SID-JM-TAN(SCA). We analyzed spectral signatures obtained from 13 different human tissue types and two different materials (gauze, instruments), collected from HS images of 100 patients during surgeries. RESULTS: The quantitative results showed the reliable performance of the different similarity measures and the proposed hybrid measures for tissue discrimination purposes. The latter produced higher discrimination values, up to 6.7 times more than the classical spectral similarity measures. Moreover, an application of the similarity measures was presented to support the annotations of the HS images. We showed that the automatic checking of tissue-annotated thyroid and colon tissues was successful in 73% and 60% of the total spectra, respectively. The hybrid measures showed the highest performance. Furthermore, the automatic labeling of wrongly annotated tissues was similar for all measures, with an accuracy of up to 90%. CONCLUSION: In future work, the proposed spectral similarity measures will be integrated with tools to support physicians in annotations and tissue labeling of HS images.

RESUMEN

Considering the high toxicity of arsenic (As), its contamination of soil represents an alarming environmental and public health issue. Existing soil heavy metal concentration estimation models based on hyperspectral data ignore the spatial nonstationarity of the relationship between the soil spectrum and heavy metal concentration. A novel model (geographically weighted eXtreme gradient boosting or GW-XGBoost model) combining geographically weighted regression (GWR) method with XGBoost algorithm was proposed. The northeast district of Beijing, China, was chosen as a case study area to assess the effectiveness of the proposed model. The GW-XGBoost model was established to estimate the As concentration based on the typical spectrum of As and the spatial correlation between the spectrum and As concentration obtained using the GWR method, and the result was compared to that obtained with the XGBoost and GWR models. The accuracy of the GW-XGBoost model was obviously better than that of the other models (R2GW-XGBoost = 0.90, R2XGBoost = 0.48, and R2GWR = 0.74). Therefore, the proposed model is reliable, as it considers the spatial correlation between the spectrum and As concentration.

Asunto(s)

RESUMEN

Hyperspectral aerial imagery is becoming increasingly available due to both technology evolution and a somewhat affordable price tag. However, selecting a proper UAV + hyperspectral sensor combo to use in specific contexts is still challenging and lacks proper documental support. While selecting an UAV is more straightforward as it mostly relates with sensor compatibility, autonomy, reliability and cost, a hyperspectral sensor has much more to be considered. This note provides an assessment of two hyperspectral sensors (push-broom and snapshot) regarding practicality and suitability, within a precision viticulture context. The aim is to provide researchers, agronomists, winegrowers and UAV pilots with dependable data collection protocols and methods, enabling them to achieve faster processing techniques and helping to integrate multiple data sources. Furthermore, both the benefits and drawbacks of using each technology within a precision viticulture context are also highlighted. Hyperspectral sensors, UAVs, flight operations, and the processing methodology for each imaging type' datasets are presented through a qualitative and quantitative analysis. For this purpose, four vineyards in two countries were selected as case studies. This supports the extrapolation of both advantages and issues related with the two types of hyperspectral sensors used, in different contexts. Sensors' performance was compared through the evaluation of field operations complexity, processing time and qualitative accuracy of the results, namely the quality of the generated hyperspectral mosaics. The results shown an overall excellent geometrical quality, with no distortions or overlapping faults for both technologies, using the proposed mosaicking process and reconstruction. By resorting to the multi-site assessment, the qualitative and quantitative exchange of information throughout the UAV hyperspectral community is facilitated. In addition, all the major benefits and drawbacks of each hyperspectral sensor regarding its operation and data features are identified. Lastly, the operational complexity in the context of precision agriculture is also presented.

Asunto(s)

RESUMEN

Estimation of the amino acid content in maize leaves is helpful for improving maize yield estimation and nitrogen use efficiency. Hyperspectral imaging can be used to obtain the physiological and biochemical parameters of maize leaves with the advantages of being rapid, non-destructive, and high throughput. This study aims to estimate the multiple amino acid contents in maize leaves using hyperspectral imaging data. Two nitrogen (N) fertilizer experiments were carried out to obtain the hyperspectral images of fresh maize leaves. The partial least squares regression (PLSR) method was used to build the estimation models of various amino acid contents by using the reflectance of all bands, sensitive band range, and sensitive bands. The models were then validated with the independent dataset. The results showed that (1) the spectral reflectance of most amino acids was more sensitive in the range of 400-717.08 nm than other bands. The estimation accuracy was better by using the reflectance of the sensitive band range than that of all bands; (2) the sensitive bands of most amino acids were in the ranges of 505.39-605 nm and 651-714 nm; and (3) among the 24 amino acids, the estimation models of the ß-aminobutyric acid, ornithine, citrulline, methionine, and histidine achieved higher accuracy than those of other amino acids, with the R 2, relative root mean square error (RE), and relative percent deviation (RPD) of the measured and estimated value of testing samples in the range of 0.84-0.96, 8.79%-19.77%, and 2.58-5.18, respectively. This study can provide a non-destructive and rapid diagnostic method for genetic sensitive analysis and variety improvement of maize.

RESUMEN

Objectives: This research aims to apply an auditory display for tumor imaging using fluorescence data, discuss its feasibility for in vivo tumor evaluation, and check its potential for assisting enhanced cancer perception. Methods: Xenografted mice underwent fluorescence imaging after an injection of cy5.5-glucose. Spectral information from the raw data was parametrized to emphasize the near-infrared fluorescence information, and the resulting parameters were mapped to control a sound synthesis engine in order to provide the auditory display. Drag−click maneuvers using in-house data navigation software-generated sound from regions of interest (ROIs) in vivo. Results: Four different representations of the auditory display were acquired per ROI: (1) audio spectrum, (2) waveform, (3) numerical signal-to-noise ratio (SNR), and (4) sound itself. SNRs were compared for statistical analysis. Compared with the no-tumor area, the tumor area produced sounds with a heterogeneous spectrum and waveform, and featured a higher SNR as well (3.63 ± 8.41 vs. 0.42 ± 0.085, p < 0.05). Sound from the tumor was perceived by the naked ear as high-timbred and unpleasant. Conclusions: By accentuating the specific tumor spectrum, auditory display of fluorescence imaging data can generate sound which helps the listener to detect and discriminate small tumorous conditions in living animals. Despite some practical limitations, it can aid in the translation of fluorescent images by facilitating information transfer to the clinician in in vivo tumor imaging.

RESUMEN

Imaging spectroscopy provides the opportunity to incorporate leaf and canopy optical data into ecological studies, but the extent to which remote sensing of vegetation can enhance the study of belowground processes is not well understood. In terrestrial systems, aboveground and belowground vegetation quantity and quality are coupled, and both influence belowground microbial processes and nutrient cycling. We hypothesized that ecosystem productivity, and the chemical, structural and phylogenetic-functional composition of plant communities would be detectable with remote sensing and could be used to predict belowground plant and soil processes in two grassland biodiversity experiments: the BioDIV experiment at Cedar Creek Ecosystem Science Reserve in Minnesota and the Wood River Nature Conservancy experiment in Nebraska. We tested whether aboveground vegetation chemistry and productivity, as detected from airborne sensors, predict soil properties, microbial processes and community composition. Imaging spectroscopy data were used to map aboveground biomass, green vegetation cover, functional traits and phylogenetic-functional community composition of vegetation. We examined the relationships between the image-derived variables and soil carbon and nitrogen concentration, microbial community composition, biomass and extracellular enzyme activity, and soil processes, including net nitrogen mineralization. In the BioDIV experiment-which has low overall diversity and productivity despite high variation in each-belowground processes were driven mainly by variation in the amount of organic matter inputs to soils. As a consequence, soil respiration, microbial biomass and enzyme activity, and fungal and bacterial composition and diversity were significantly predicted by remotely sensed vegetation cover and biomass. In contrast, at Wood River-where plant diversity and productivity were consistently higher-belowground processes were driven mainly by variation in the quality of aboveground inputs to soils. Consequently, remotely sensed functional, chemical and phylogenetic composition of vegetation predicted belowground extracellular enzyme activity, microbial biomass, and net nitrogen mineralization rates but aboveground biomass (or cover) did not. The contrasting associations between the quantity (productivity) and quality (composition) of aboveground inputs with belowground soil attributes provide a basis for using imaging spectroscopy to understand belowground processes across productivity gradients in grassland systems. However, a mechanistic understanding of how above and belowground components interact among different ecosystems remains critical to extending these results broadly.

RESUMEN

Mining and smelting releases toxic contaminants such as zinc (Zn), lead (Pb) or cadmium (Cd) into the soil thereby poisoning it and rendering it unproductive. Remotely alternatives have been widely employed in the attempt of estimating heavy metal content within soils. The present study provides a methodological approach based on VNIR-SWIR field hyperspectral data and multispectral Landsat OLI 8 imageries for the prediction and mapping of Pb, Zn and Cd heavy metal contents around the abandoned Jebel Ressas mine site in Northern Tunisia. Thus, eighty-seven soil and tailing samples were collected from the study site and VNIR-SWIR field reflectances were measured on the same collection points, as well. All samples were analysed by atomic absorption for the estimation of heavy metal concentrations. The partial least squares regression PLSR was conducted considering the measured heavy metal concentrations and using multi-scale data: VNIR-SWIR field hyperspectral data and multispectral Landsat OLI 8 imagery. Standard normal variable (SNV) and multiple scatter correction (MSC) preprocessing methods were applied for further mapping improvement. Thus, this work aims to automate the estimation of the heavy metal contents in contaminated soils, by carrying out: a modeling approach based on the PLSR using VNIR-SWIR field hyperspectral data, ii) the mapping of Pb and Zn contents thanks to the exploitation of Landsat OLI8 multispectral imagery and iii) the application of both MSC and SNV preprocessing methods to optimize the performance of the developed models, when using such spectrally and spatially degraded data.

RESUMEN

Estimating soil properties is important for maximizing the production of crops in sustainable agriculture. The hyperspectral data next input depends upon the previous one, and the current techniques do not take advantage of this sequential nature of hyperspectral signatures. The variants of RNN can learn the short-term and long-term dependencies between data. This paper proposes a deep learning hybrid framework for quantifying the soil minerals like Clay, CEC, pH of H2O, Nitrogen, Organic Carbon, Sand of European Union from the LUCAS library. The hyperspectral signatures contain the data in the range of 400-2500 nm captured from the FOSS spectroscope in the laboratory. As hyperspectral data is high dimensional, Principal Component Analysis and Locality Preserving Projections are utilized to form the hybrid features, which have low dimensions containing the local and global information of the original dataset. These hybrid features are passed on to Long Short Term Memory Networks, a deep learning framework for building an effective prediction model. The effectiveness of the prepared models is demonstrated by comparing it to existing state-of-the-art techniques.

Asunto(s)

RESUMEN

Classical approaches in cluster analysis are typically based on a feature space analysis. However, many applications lead to datasets with additional spatial information and a ground truth with spatially coherent classes, which will not necessarily be reconstructed well by standard clustering methods. Motivated by applications in hyperspectral imaging, we introduce in this work clustering models based on Orthogonal Nonnegative Matrix Factorization (ONMF), which include an additional Total Variation (TV) regularization procedure on the cluster membership matrix to enforce the needed spatial coherence in the clusters. We propose several approaches with different optimization techniques, where the TV regularization is either performed as a subsequent post-processing step or included into the clustering algorithm. Finally, we provide a numerical evaluation of 12 different TV regularized ONMF methods on a hyperspectral dataset obtained from a matrix-assisted laser desorption/ionization imaging measurement, which leads to significantly better clustering results compared to classical clustering models.