Um novo método para auferição de características de cascas de ovos utilizando processamento de imagens

Abstract

The geometric and visual characteristics of bird egg shells have been objects of study for more than half a century, as they allow us to determine the healthiness of the egg and its suitability for the various stages of the production chain. This work proposes a nondestructive and low-cost method (named here as EggScale) for measuring the dimensional and color characteristics of eggshells, both factors involved in determining the hatchability of these eggs. A capture apparatus was designed to acquire images under stable lighting conditions and, from the processing of the images obtained, it was possible to accurately extract the length and maximum width measurements of the eggs (average error of 1.64mm and 1.83mm , respectively) using object detection techniques and identification of their edge points. The volume was estimated from the calculation of cone frustums based on the maximum length and width measurements identified through image processing. Good accuracy rates were achieved, with an average error rate of 4.27ml with a standard deviation of 2.16 in the worst case, even achieving better results than those obtained by the method established by Narushin (NARUSHIN; ROMANOV; GRIFFIN, 2021b) and used as a basis for evaluating results. A method for analyzing eggshell color was also proposed. This analysis proved more challenging, however, as the data obtained indicated low reliability in establishing a reference standard, given the high level of subjectivity involved in color perception by human evaluators.