Métodos de identificação para separação de sementes de arroz quanto à diferença de pigmentação, dimensões e fissuras por meio do processamento de imagens

Abstract

For fissure evaluation, samples of three lots of rice seeds were used, a lot containing moist seed not subjected to drying (peeled seed) and two lots submitted to drying, a seed with peel and another containing seedless seed, each Sample containing 100 seeds. Images in X-ray and RGB formats were provided in the sequence processed in the ImageJ - FIJI software and introduced in the machine learning software, where they were pre-processed using the appropriate filters and then classified through the J48 and LDA classifiers. X-ray images using differentiated equipment allow identifying cracks in rice seeds by using image processing techniques and the LDA classifier. In addition, the capture of images by RGB is a viable alternative. The use of filters individually or in combination may consist of an adequate alternative for rice seed classification. For red and black rice seed separation of white rice seeds, the most suitable color component is the scale of red. On the other hand, the most efficient response is obtained with the blue scale for black and red rice seeds. It is possible to use image processing techniques for the differences in seeds, and the length and width ratios are the most promising to achieve greater process efficiency, accurately from 72 and 73%, respectively