Development of forestry seedling inspection & quality control machine

• Grant number: 15/08706-5
• Support type: Research Grants - Innovative Research in Small Business - PIPE
• Duration: June 01, 2016 - May 31, 2018
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Cooperation agreement: FINEP - PIPE/PAPPE Grant
• Principal researcher: Fernando Antonio Torres Velloso da Silva Neto
• Grantee: Fernando Antonio Torres Velloso da Silva Neto
• Company: Mvisia Comércio de Equipamentos Eletrônicos Inovadores Ltda
• City: São Paulo
• Pesquisadores principais: Nelson de Moura Martins Gomes
• Assoc. researchers: Fernando Paes Lopes ; Laio Burim Vilas Bôas ; Marcelo Li Koga

Abstract

In the productive process of eucalyptus in commercial scale, one of the difficulties growers encounter is to ensure the good development of the plants, which may have diseases or grow non-homogeneously. In this scenario, the seedlings that are used are very important because we can estimate the growing potential and identify some diseases through a visual inspection on them. Therefore, classifying and sorting the seedlings prior to planting improves the quality of the production, avoiding losses. In order to achieve the automation of this task, we can then build a machine for inspection and quality control of eucalyptus seedlings. Given the visual aspect of the seedlings sorting, we can use a computer vision system. We extract several parameters of the seedlings via image processing and use them as input to a classifier (of quality levels). Using examples of seedlings previously classified by a specialist, we can built a classifier using machine learning. The machine must also be able to manipulate the seedlings' tubes with electromechanical components, removing them from trays, positioning them for image capture and sorting them on the several classes, putting them back on new trays. The objective of this project is thereby develop the necessary means and methods for achieving automatic classification of eucalyptus' seedlings in several quality levels and build a prototype of a machine to perform this task. Trays with seedlings in tubes are the input and five reorganized trays, each one of which with just seedlings of a single class, are the output. By the end of this research project, we expect to possess the prototype of a machine capable of correctly classifying eucalyptus seedlings. In future work, this prototype will become a product, providing an increase in productivity, quality and profitability for national growers. (AU)