Estimation of surface's fruit, using 3D geometric models and geometric projection method, applied on digital pictures.
Loera Alvarado G., Chavez Franco S. H., Carrillo Salazar J. A., Valdez Carrasco J. M.
Author Affiliation: Programa de Fruticultura, Colegio de Postgraduados, Carretera México--Texcoco, km 36.5, C.P. 56230, Montecillo, Texcoco, Estado de México, Mexico.
Revista Iberoamericana de Tecnologia Postcosecha 15 : 119-126
Abstract : Two techniques based on digital image processing were developed; to generate a three-dimensional geometric model in CAD and other one to generate a geometric projection in three fruit species (mango, orange and pear). Four images were taken; one of the widest side (0°) and the others rotating fruit 90°. The models were generated dividing each image in 8, 16 and 32 slices (sections). The geometric projection was generated by correcting its vertical and horizontal curvature by calculating their arc length pixel by pixel or each two pixels. The effect of using several sizes of the images (resolution) at 50%, 100% and 200% of its original size was evaluated. Fruit's skin was removed and scanned to calculate its area. Volume of fruit was measured by water displacement using the geometrical model. Data were analyzed using a randomized block design for each king of fruit separately. In relation to volume; there was only significant difference when modeled whit 8 divisions (slices). Area showed that geometric modeling were not significant in all of them. Not significance was found; in mango, using fruit images at 200% every two pixels, in orange images at 100% and 200% per pixel and in pear images at 100% and 200% pixel by pixel and 50% every two pixels. Volume error was 3% in the three species. In relation to surface area the smallest error was obtained in models using 32 divisions (slices) with the following values, 1.4%, 0.6% and 1.3% (R2 0.98, 0.99 and 0.88), while the geometric projection showed 1.6%, 1.1% and 5.4% (R2 0.97, 0.96 and 0.07) for mango, orange and pear respectively.