Perfil de calidad de cacao utilizando espectroscopía de infrarrojo cercano portable, desafíos para diferenciación geográfica
Fecha
2024
Autores
Quesada Arguedas, Paola
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional, Costa Rica
Resumen
La combinación de curvas de calibración generadas por espectroscopía de infrarrojo cercano (NIR) con aprendizaje profundo supone una oportunidad para crear métodos de discriminación de la calidad y origen del cacao, apoyar estrategias de valorización territorial y trazabilidad de cacaos diferenciados. Validar la aplicabilidad de un método de discriminación de cacaos según su origen geográfico, mediante técnicas de espectroscopía NIR portátil y aprendizaje profundo. Se recolectaron 176 muestras de granos de cacao fermentados y secos de diferentes regiones de Costa Rica, utilizando 72 muestras para calibración y 106 para predicción. Las muestras se procesaron para análisis proximal, acidez titulable y compuestos fenólicos. Se empleó un espectrofotómetro NIR para recopilar datos espectrales (400-1700 nm). Se realizó un preprocesamiento de la información espectral y se desarrollaron modelos de regresión (XLS, PLS, PCR) para predecir características químicas. Para el modelo de clasificación geográfica se eliminó el ruido de los espectros y se realizó un análisis de conglomerados utilizando la distancia Gower y el método de agrupamiento Ward con los componentes obtenidos en Análisis de Componentes Principales (ACP). La grasa destacó como el principal componente (>39,67%). El análisis espectral reveló la capacidad del NIR para diferenciar cacao según el grado de fermentación y contenido de compuestos fenólicos. El modelo XLS demostró la mejor capacidad predictiva para propiedades químicas. El agrupamiento por origen geográfico reflejó cuatro grupos influenciados mayormente por propiedades químicas asociadas a prácticas poscosecha. Conclusión. El modelo de regresión lineal utilizado mostró superioridad en la predicción de características químicas proximales. Se señala que la limitada diversidad genética y prácticas poscosecha estandarizadas pueden reducir la variabilidad de calidad asociada al origen geográfico, limitando la utilidad del NIR en la identificación de origen y trazabilidad. Se sugiere explorar espectros más amplios y equipos adicionales para análisis multivariados avanzados.
The combination of calibration curves generated by near-infrared spectroscopy (NIR) with deep learning presents an opportunity to create methods for discriminating the quality and origin of cocoa, supporting territorial valorization strategies and traceability of differentiated cocoas. To validate the applicability of a cocoa discrimination method based on geographical origin, using portable NIR spectroscopy and deep learning techniques. 176 samples of fermented and dried cocoa beans from different regions of Costa Rica were collected, using 72 samples for calibration and 106 for prediction. The samples underwent proximal analysis, titratable acidity, and phenolic compound analysis. A NIR spectrophotometer was used to collect spectral data (400-1700 nm). Spectral information was pre-processed, and regression models (XLS, PLS, PCR) were developed to predict chemical properties. For the geographical classification model, noise was removed from the spectra, and cluster analysis was performed using Gower distance and Ward clustering with components obtained in Principal Component Analysis (PCA). Fat stood out as the main component (>39.67%). Spectral analysis revealed the ability of NIR to differentiate cocoa based on fermentation level and phenolic compound content. The XLS model demonstrated the best predictive capacity for chemical properties. Geographical clustering reflected four groups mainly influenced by chemical properties associated with post-harvest practices. The linear regression model used showed superiority in predicting proximal chemical characteristics. It is noted that limited genetic diversity and standardized post-harvest practices may reduce the variability of quality associated with geographical origin, limiting the utility of NIR in identifying origin and traceability. Exploring broader spectra and additional equipment for advanced multivariate analysis is suggested.
The combination of calibration curves generated by near-infrared spectroscopy (NIR) with deep learning presents an opportunity to create methods for discriminating the quality and origin of cocoa, supporting territorial valorization strategies and traceability of differentiated cocoas. To validate the applicability of a cocoa discrimination method based on geographical origin, using portable NIR spectroscopy and deep learning techniques. 176 samples of fermented and dried cocoa beans from different regions of Costa Rica were collected, using 72 samples for calibration and 106 for prediction. The samples underwent proximal analysis, titratable acidity, and phenolic compound analysis. A NIR spectrophotometer was used to collect spectral data (400-1700 nm). Spectral information was pre-processed, and regression models (XLS, PLS, PCR) were developed to predict chemical properties. For the geographical classification model, noise was removed from the spectra, and cluster analysis was performed using Gower distance and Ward clustering with components obtained in Principal Component Analysis (PCA). Fat stood out as the main component (>39.67%). Spectral analysis revealed the ability of NIR to differentiate cocoa based on fermentation level and phenolic compound content. The XLS model demonstrated the best predictive capacity for chemical properties. Geographical clustering reflected four groups mainly influenced by chemical properties associated with post-harvest practices. The linear regression model used showed superiority in predicting proximal chemical characteristics. It is noted that limited genetic diversity and standardized post-harvest practices may reduce the variability of quality associated with geographical origin, limiting the utility of NIR in identifying origin and traceability. Exploring broader spectra and additional equipment for advanced multivariate analysis is suggested.
Descripción
Licenciatura en Ingeniería en Agronomía con la modalidad de Artículo Científico
Palabras clave
ESPECTROSCOPIA, CACAO, SEMILLA, ANÁLISIS ESPECTROSCÓPICO, CALIDAD, THEOBROMA CACAO, PRODUCCIÓN, CONTROL DE CALIDAD, COSECHA, COSTA RICA