[329698] Artigo – CBMO – Congresso Brasileiro de Metrologia Óptica

Resumo: Law 9,972 of 2000 instituted the classification of vegetable products, byproducts, and waste of economic value, making their classification mandatory when intended for human consumption, in purchase and sale operations, and at ports and airports when destined for export. The classification of these vegetables consists of separating the product into different categories of quality according to their peculiarities, forming homogeneous batches, as well as determining the quality based on official standards. Among the quality standards to be observed in this classification process are shape, size, weight, defects in the skin, and color, which will be the focus of this work. Currently, much of this classification process is carried out manually, based on the knowledge and experience of operators. With a focus on automating this type of process, studies in the field of computer vision are being developed with the aim of autonomously reproducing the classification of fruits and vegetables usually carried out by trained operators. However, in various applications in the current literature, digital cameras are used to extract quantitative measurements. These pieces of information require metrological analyses like any other conventional measurement system. In this sense, the present paper proposes a study to evaluate the errors in the application of a digital camera for color measurement. Results will be shown from tests comparing numerical aperture, exposure time, and ISO parameters with the results of a colorimeter using a color chart as a measurand. The goal is to evaluate the accuracy and estimate the error curves of a conventional sensor when applied to color metrology. This study is an important step in ensuring the reliability of color measurements using digital cameras for automatic fruit and vegetable classification systems.