Development and validation of multispectral imaging methods to analyze the quality of maize seeds

• Grant number: 22/11706-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: September 01, 2022
• End date: August 31, 2023
• Field of knowledge: Agronomical Sciences - Agronomy - Crop Science
• Principal Investigator: Clíssia Barboza Mastrangelo
• Grantee: Natália Chittolina
• Host Institution: Centro de Energia Nuclear na Agricultura (CENA). Universidade de São Paulo (USP). Piracicaba , SP, Brazil
• Associated research grant: 17/15220-7 - Non-destructive image analysis methods for seed quality evaluation, AP.JP

Abstract

Recent advances in multispectral imaging technology and the expansion of knowledge on electromagnetic properties of different seed tissues have allowed the development of new imaging methods, with fast and non-destructive measurements of seed quality. From the creation of machine learning models, physical, chemical, physiological and health attributes of seed lots can be objectively analyzed with real-time control. In recent years, our team has accumulated fundamental knowledge about the use of modern optical imaging resources for automating seed quality testing. Therefore, we propose to use such knowledge to create machine learning models based on innovative artificial intelligence algorithms for the development and validation of multispectral imaging methods with potential for maize seed classification. The multispectral imaging technology will be employed in different commercial lots of maize seeds from different hybrids, using the VideometerLab4® instrument (Videometer A/S, Herlev, Denmark) that allows a simultaneous extraction of texture, color, and chemical composition features of the seeds. Machine learning models will be developed for seed classification, and then the features extracted from the optical images will be correlated to germination and vigor tests (first germination count, cold test, seedling emergence, and emergence speed index - ESI), in addition to X-ray tests. We expect to validate machine learning models for fast, accurate, and non-invasive evaluation of maize seed lots from robust multispectral imaging systems, contributing to the strengthening and expansion of innovative approaches involving recent imaging technologies.(AU)