Enzymatic Browning in Fresh-Cut Apple Slices Measured byudDifferent Kinds of Image Algorithms

摘要

ResumenudEl objetivo final es el desarrollo de un sistema di visión multiespectral que permita asignar manzanas cortadas a clases de distinto nivel de pardeamiento. Se ha analizado un total de 240 imágenes IRRB y RGB, correspondientes a 240 gajos de manzanas de la variedad ‘Granny Smith’ (120 gajos = Set 1; 120 gajos = Set 2). Se analizaron 24 gajos por día: a tiempo cero y después de 1, 3, 7 y 9 días de almacenamiento a 7,5ºC. A las imágenes virtuales obtenidas como combinación del canal rojo y azul (B/R, R-B y (R-B)/(R+B)) se aplicó un procedimiento de clasificación no supervisada que, en todo los casos, generó tres clases de referencia. A la segunda serie de muestras (Set 2), sometidas los mismos tratamientos, se aplicó una validación externa, obteniendo un alto porcentaje de muestras correctamente clasificadas. La clasificación de las cámaras IRRB y RGB se evaluó de acuerdo a parámetros colorimétricos y sensoriales y las imágenes virtuales (R-B)/(R+B) y B/R mostraron la mejor sensibilidad para reflejar el cambio de color asociado con el pardeamiento.ududAbstractudThe main objective of this study was to develop a vision system able to classify fresh-cut apple slices according to the development of enzymatic browning. The experiment was carried out on ‘Granny Smith’ apple slices stored at 7.5°C (Set 1 = 120 n). Twenty-four samples were analyzed per day: at zero time and after storage for 1, 3, 7 and 9 days. Digital images were acquired by employing an IRRB camera and by employing a cheaper vision system, consisting in a RGB digital camera. A classification procedure was applied to the histograms of the following virtual images, acquired by the IRRB and by the RGB camera: (R-B)/(R+B), R–B and B/R. In all cases, a non-supervised classification procedure was able to generate three image-based browning reference classes. An internal and an external validation (Set 2 = 120 n) were carried out, with a high percentage of corrected classified samples. The camera classification was evaluated according to reference parameters (colorimetric and sensorial measurements) and the best results were obtained with the (R-B)/(R+B) and B/R virtual images.