SISTEMA MULTIESPECTRAL COM INTELIGÊNCIA ARTIFICIAL PARA ANÁLISE NÃO DESTRUTIVA DE SEMENTES NATIVAS DA AMAZÔNIA LEGAL
SISTEMA MULTIESPECTRAL COM INTELIGÊNCIA ARTIFICIAL PARA ANÁLISE NÃO DESTRUTIVA DE SEMENTES NATIVAS DA AMAZÔNIA LEGAL
-
DOI: https://doi.org/10.22533/at.ed.932112624042
-
Palavras-chave: Sementes nativas. Visão multiespectral. Inteligência artificial. Análise não destrutiva. Amazônia Legal.
-
Keywords: Native seeds. Multispectral vision. Artificial intelligence. Non-destructive analysis. Brazilian Legal Amazon.
-
Abstract: This article presents the development of a multispectral system with artificial intelligence for the non-destructive analysis of native seeds from the Brazilian Legal Amazon. The research adopted an applied, exploratory, and technological approach, focused on the construction and operational validation of a prototype capable of integrating optical image acquisition, controlled illumination, batch-based data organization, image processing, and data preparation for computational modeling. The solution was structured with sensors/cameras, a conveyor belt, a local computational application, and post-processing routines, using spectral bands of 650 nm, 685 nm, and 690 nm. The results demonstrated positive performance in the acquisition, storage, and organization of multispectral blocks, as well as in the generation of masks, normalization, edge maps, and radiometric diagnosis. The technical evaluation indicated greater initial usefulness of the 650 nm and 685 nm bands, while the 690 nm band required complementary illumination calibration. The study concludes that the proposed solution presents technical feasibility to support the screening, traceability, and preliminary analysis of native seeds, providing a structured basis for labeling, validation with germination and vigor parameters, and subsequent application of artificial intelligence models.
- David Lopes Maciel
- Walter Pinheiro Rodrigues
- Brício Rocha Borges de Almeida
- Fabrício Moraes de Almeida