Color chart development by computer vision for flue-cured tobacco leaf

Abstract

Currently, the quality of flue-cured tobacco leaves is evaluated manually, which relies on subjective experience and inevitably affected by personal, physical, and environmental factors. However, the subjective evaluation fails to meet the automatic and precise requirements of tobacco production. The quality of tobacco leaves is affected by a variety of factors, such as color, oil content, maturity, and surface texture, among which color is one of the most important factors. Color evaluation is critical for quality management in the agricultural field. However, there is no specific standard color chart for flue-cured tobacco leaves, and there are few studies focusing the development of such a color chart. In this work, a framework for the development of a color chart was established by computer vision techniques and it was applied to flue-cured tobacco leaves. The color chart system (CCS) consisted of data acquisition, color representation for a single leaf, and the development of the color chart for flue-cured tobacco leaves. Firstly, an acquisition device was developed to collect digital color images of flue-cured tobacco leaves, which was equipped with CCD sensors that detect color information accurately. Secondly, a proportional threshold method was proposed to represent a feature color from the acquired images by taking into consideration the overall color information of a single leaf. Finally, color discrimination techniques were used to create digital color charts depicting synthetic, standard-color tobacco leaves at various degrees. The color charts established in this work faithfully express the flue-cured tobacco leaves color information, and the framework of CCS based on computer vision can be applied to other agricultural situations where color estimation is required.