ABSTRACT

Sorting seeds is important in agriculture for both productivity and commercial reasons. Low-quality seeds may result in poor plant development, disease, and poor crop yields. A quick identification and classification approach for maize seeds is developed in this work using machine learning and machine vision. The traditional experiments’ knowledge cannot reduce agriculture’s present adverse effects on the biosphere. There is an increase in the gap between the slowly expanding knowledge base and the adverse impacts of the environment. Crop seed quality can be significantly indicated by the seed purity. Also, in the modern agricultural industry, maize is a significant crop which is having worldwide production greater than 40%. All over the world, maize is among the significantly cultivated grains. In the context of genetic programs, modern crop improvement and advanced maize breeding, the most significant technique is doubled-haploid as with the help of this technique breeding efficiency is increased as well as breeding period is shortened. The study aims in examining the ML (machine learning) approach’ feasibility in various types of maize seeds’ classification. The digital images (DI) of seeds of 6 maize varieties were ICI 339, Neelam Makkai, Pioneer P-1429, Kashmiri Makkai, Sygenta ST-6142, and Desi Makkai. The classification outcomes meet the demands of both producers and consumers.

Keywords: Seed, Environment, Machine learning, Agriculture, Virtual Reality, Quality, Random Forest, Confusion Matrix.