RESUMO

Large amount of information in hyperspectral images (HSI) generally makes their analysis (e.g., principal component analysis, PCA) time consuming and often requires a lot of random access memory (RAM) and high computing power. This is particularly problematic for analysis of large images, containing millions of pixels, which can be created by augmenting series of single images (e.g., in time series analysis). This tutorial explores how data reduction can be used to analyze time series hyperspectral images much faster without losing crucial analytical information. Two of the most common data reduction methods have been chosen from the recent research. The first one uses a simple randomization method called randomized sub-sampling PCA (RSPCA). The second implies a more robust randomization method based on local-rank approximations (rPCA). This manuscript exposes the major benefits and drawbacks of both methods with the spirit of being as didactical as possible for a reader. A comprehensive comparison is made considering the amount of information retained by the PCA models at different compression degrees and the performance time. Extrapolation is also made to the case where the effect of time and any other factor are to be studied simultaneously.

Assuntos

Distribuição Aleatória , Análise de Componente Principal

RESUMO

Remote identification of illegal plantations of Cannabis sativa Linnaeus is an important task for the Brazilian Federal Police. The current analytical methodology is expensive and strongly dependent on the expertise of the forensic investigator. A faster and cheaper methodology based on automatic methods can be useful for the detection and identification of Cannabis sativa L. in a reliable and objective manner. In this work, the high potential of Near Infrared Hyperspectral Imaging (HSI-NIR) combined with machine learning is demonstrated for supervised detection and classification of Cannabis sativa L. This plant, together with other plants commonly found in the surroundings of illegal plantations and soil, were directly collected from an illegal plantation. Due to the high correlation of the NIR spectra, sparse Principal Component Analysis (sPCA) was implemented to select the most important wavelengths for identifying Cannabis sativa L. One class Soft Independent Class Analogy model (SIMCA) was built, considering just the spectral variables selected by sPCA. Sensitivity and specificity values of 89.45% and 97.60% were, respectively, obtained for an external validation set subjected to the s-SIMCA. The results proved the reliability of a methodology based on NIR hyperspectral cameras to detect and identify Cannabis sativa L., with only four spectral bands, showing the potential of this methodology to be implemented in low-cost airborne devices.

Assuntos

Cannabis/química , Imageamento Hiperespectral/métodos , Imageamento Hiperespectral/estatística & dados numéricos , Aprendizado de Máquina , Brasil , Quimioinformática , Estudos de Viabilidade , Folhas de Planta/química , Análise de Componente Principal , Reprodutibilidade dos Testes , Sensibilidade e Especificidade , Espectroscopia de Luz Próxima ao Infravermelho/métodos

RESUMO

Industry requires non-destructive real-time methods for quality control of meat in order to improve production efficiency and meet consumer expectations. Near Infrared Hyperspectral Images were used for tenderness evaluation of Nellore beef and the construction of tenderness distribution maps. To investigate whether the selection of the region of interest (ROI) in the image at the exact location where the shear force core was collected improves tenderness prediction and classification models, 50 samples from Longissimus muscle were imaged (1000-2500 nm) and shear force were measured (Warner-Bratzler). The data were analyzed by chemometric techniques (Partial Least Squares together with discriminant analysis - PLS-DA). Classification models using local ROI presented better performance than the ROI models of the whole sample (external validation sensitivity for the tough class = 33% and 70%, respectively), but none could be considered as successful model. However, the more general model had better performance in the tenderness distribution maps, with 72% of predicted images correctly classified.

Assuntos

Inspeção de Alimentos/métodos , Qualidade dos Alimentos , Carne/análise , Modelos Biológicos , Músculos Paraespinais/química , Matadouros , Animais , Animais Endogâmicos , Brasil , Calibragem , Bovinos , Análise Discriminante , Estudos de Viabilidade , Inspeção de Alimentos/instrumentação , Análise dos Mínimos Quadrados , Análise de Componente Principal , Reprodutibilidade dos Testes , Resistência ao Cisalhamento , Espectroscopia de Luz Próxima ao Infravermelho

RESUMO

The interest in performing in field measures using portable instruments is growing increasingly. Calibration transfer techniques can be used to enable models, predicted values or spectra obtained in a benchtop instrument be used in portable instrument, saving money and time required for a complete recalibration. Most of the calibration transfer methods require a set of transfer samples which spectra have to be acquired in both spectrometers. The present work evaluates the use of virtual standards as transfer samples in the reverse standardization (RS) method in order to standardize very dissimilar spectral responses of fuel samples (gasoline and biodiesel blends) from a high-resolution benchtop Frontier FT-NIR (PerkinElmer) spectrometer and a handheld MicroNIR™1700 (JDSU). These virtual standards can be created by mathematically mixing spectra from the pure solvents present in gasoline or diesel/biodiesel (D/B) blends, to avoid volatilization and changes in the composition of the compounds during storage and/or transportation of the real transfer fuel samples. Virtual standards were created using ten and five pure solvents for gasoline and D/B blends, respectively. Partial least squares regression (PLS) models were built for five quality parameters of gasoline (distillation temperatures at 10%, 50%, 90% and final boiling point (FBP) volume recovered and density) and one of D/B blends (biodiesel content). The RMSEP values obtained after the standardization approaches were equivalent to the reproducibility of the reference methods, except for density and biodiesel content parameters obtained for the virtual samples standardization approach. RS procedure provided promising results showing that it is possible to transfer gasoline or D/B blend spectra acquired with a high-resolution benchtop instrument to the handheld MicroNIR using virtual standards as transfer samples.