Determinación de la concentración de minerales en malanga morada (Colocasia esculenta (L.) Schott) mediante el uso de la espectroscopia VIS/NIR
Archivos
Fecha
2016-06-03
Autores
Sánchez Valle, Sandra Adyenne
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Central "Marta Abreu" de Las Villas. Facultad de Ciencias Agropecuarias. Departamento de Ingeniería Agrícola.
Resumen
La investigación se desarrolló con el fin de determinar la concentración de
minerales en tres clones de malanga morada mediante el uso de la técnica de
espectroscopia Vis/NIR, para la obtención de resultados más precisos con menor
gasto de recursos y tiempo. Los espectros de las muestras de malanga fueron
obtenidos en condiciones de laboratorio, mediante la utilización de un
espectrofotómetro Vis/NIR portátil en el modo de reflectancia, ajustándose un
rango de longitud de onda entre 379 – 1700 nm. Los modelos matemáticos se
desarrollaron en Matlab 7.9 mediante técnicas multivariadas de regresión de
mínimos cuadrados parciales (PLS). Esta técnica facilitó la correlación entre los
espectros de reflectancia y las concentraciones de Ca, Mg, Mn, Cu, Na, K, Fe, Zn
determinados por métodos convencionales. El método de pre-tratamiento
espectral utilizado fue el de Corrección de Señal Multiplicativa (MSC), Primera
Derivada de Savitzky – Golay, Suavizado y Centrado medio. Se utilizó la técnica
de validación cruzada (Leave One Out), para optimizar la complejidad de los
modelos. Los mejores resultados en la predicción de las concentraciones de
minerales en el clon 204 se obtuvieron en Fe (R
2
= 0.92) y Mn (R
2
= 0.99), en el
clon 206 los minerales mostraron un coeficiente R
2
superior a 0.80, excepto en Ca
(R
2
= 0.70), y en el clon 207 los coeficientes R
2
fueron superiores a 0.90 en todos
los minerales analizados, excepto en Cu (R
2
= 0.81).
The research work was carried out in order to determine the mineral concentration in three clones of purple taro using the Vis/NIR spectroscopy technique for obtaining accurate results with low resources and time effective. The sample spectra were obtained in laboratory conditions by means of a portable Vis/NIR spectrophotometer in reflectance mode, with a wavelength range between 379 – 1700 nm. The mathematical models were developed in Matlab 7.9 by means of multivariate techniques of Partial Least Square Regression (PLS). This technique allowed the correlation between the reflectance spectra and the concentrations of Ca, Mg, Mn, Cu, Na, K, Fe, and Zn determined with conventional methods. The spectral pre-treatment method was the Multiplicative Signal Correction (MSC), First Derivative of Savitzky – Golay, Smoothing and Mean Center. The cross-validation technique (Leave One Out) was used to optimize the models complexity. The best prediction results of mineral concentrations in clone 204 were obtained in Fe (R 2 = 0.92) and Mn (R 2 = 0.99); in clone 206 all minerals showed a R 2 coefficient highest than 0.80, with the exception of Ca (R 2 = 0.70); in clon 207 the R 2 coefficients were highest than 0.90 in all minerals analyzed, with the exception of Cu (R 2 = 0.81).
The research work was carried out in order to determine the mineral concentration in three clones of purple taro using the Vis/NIR spectroscopy technique for obtaining accurate results with low resources and time effective. The sample spectra were obtained in laboratory conditions by means of a portable Vis/NIR spectrophotometer in reflectance mode, with a wavelength range between 379 – 1700 nm. The mathematical models were developed in Matlab 7.9 by means of multivariate techniques of Partial Least Square Regression (PLS). This technique allowed the correlation between the reflectance spectra and the concentrations of Ca, Mg, Mn, Cu, Na, K, Fe, and Zn determined with conventional methods. The spectral pre-treatment method was the Multiplicative Signal Correction (MSC), First Derivative of Savitzky – Golay, Smoothing and Mean Center. The cross-validation technique (Leave One Out) was used to optimize the models complexity. The best prediction results of mineral concentrations in clone 204 were obtained in Fe (R 2 = 0.92) and Mn (R 2 = 0.99); in clone 206 all minerals showed a R 2 coefficient highest than 0.80, with the exception of Ca (R 2 = 0.70); in clon 207 the R 2 coefficients were highest than 0.90 in all minerals analyzed, with the exception of Cu (R 2 = 0.81).
Descripción
Palabras clave
Malanga morada, Espectroscopia Vis/NIR