Análises de imagem na seleção de linhagens de arroz para qualidade de grãos

Abstract

Rice (Oryza Sativa) has been regarded as one of the most important species over time because it is a staple in the daily diet of thousands of people around the world. Along with corn and wheat crops it represents 42.5% of human energy supply. As it is a product with great relevance in the population's basic diet, the consumer's preference for the characteristics of the grain is a determining factor at the time of marketing rice. The production of whole, translucent, long and fine grains, with high quality sensory and culinary characteristics, are the main attributes of consumer preference, a fact that leaves the market strongly segmented, as preferences vary with the consuming regions, making improvement aimed at grain quality challenging. The measurement of phenotypic characteristics for grain quality is laborious, requires time and manpower, which is why it is important to develop new tools that allow speeding up evaluations in order to obtain an efficient breeding program. Image analysis has proved to be an interesting tool in rice breeding programs, bringing great advantages to breeders, generating great reliability when analyzing phenotypic data quickly, automatically and efficiently. Thus, in the present study, was validate, through the image selection methodology, the efficiency of selection of a superior improvement model to complement the maintenance of cultivars with high yield and physical grain quality, in the UFLA upland genetic improvement program. The field experiments were carried out, both in the 2018/19 and 2019/20 harvests. There was a total of 20 rice lines acquired from the Cultivation and Use Value (VCU) of the 2018/19 harvest, 88 progenies from the Observation Assay and 36 lines from the Preliminary Assay of the Upland Rice Genetic Improvement Program. The seeds of the genotypes were sown in experimental plots and, later, collected to obtain representative samples of the grains. The characteristics evaluated were: % chalkiness grain and weight of 1000 grains, both through phenotypic analysis and through the use of image analysis, using the GroudEye equipment. After data analysis, two methodologies for high-yield phenotyping were correlated for the characteristics of thousand-grain weight and gypsum percentage. The results showed positive values (0.70) for the gypsum characteristic and (0.89) for the thousand-grain weight characteristic, demonstrating that the image analysis methodology is efficient. Therefore, a new proposal for phenotypic evaluation of grain quality, helping the indirect selection of superior genotypes in the upland rice genetic improvement program.