DIGITAL IMAGE PROCESSING FOR ENDOSPERM TEXTURE ESTIMATION IN MAIZE LINES
Article Sidebar
Main Article Content
Abstract
In order to select for endosperm texture in maize lines, using the percentages of horny and floury endosperm in grains as parameters,
a technique that allows the estimation of these structural components in a fast and reliable way, was evaluated. This technique is based on
digital image processing, known as “threshold”, applied to images in a gray scale and afterwards transformed to a binary scale to generate
a contrast between background and objects. Its application on 22 lines of maize was compared to the technique of texture by dissection,
adopted by the industry of the “nixtamalized” maize flour. It was possible to evaluate grains with different horny and floury endosperm
proportions, ranging from crystalline to completely floury. The technique of image processing also allows the estimation of embryo and
pedicel percentages. The results indicated that percentages of horny and floury endosperm, obtained by image processing, were overestimated by 4 and 8 % respectively, as compared with those of texture by dissection; although there were not statistically differences between them.