Publicação
Automated Detection of Contaminants: Using Computer Vision for Food Quality Control
| Resumo: | The presence of foreign objects in food processing lines poses significant risks to consumer safety, product quality, and regulatory compliance. Traditional inspection methods such as manual visual checks or metal detectors are limited in their ability to detect non-metallic con- taminants like plastic, wood, or glass, especially when mixed with vegetable products. In re- sponse to this challenge, this work proposes the development of a real-time, computer vision system capable of detecting foreign materials on conveyor belts in a frozen vegetable factory. The proposed approach relies on RGB image acquisition and deep learning models, particu- larly object detection architectures such as YOLO (You Only Look Once), enhanced with at- tention mechanisms to improve small object recall. This document includes an overview of classical and modern vision techniques, the selection of suitable tools and models, and a defined work plan for implementation, calibration, testing and results. The expected outcome is a lightweight, accurate detection model integrated into a prototype system capable of operating in real-time under industrial conditions. This work aims to contribute to food safety automation in compliance with EU standards and to demon- strate the viability of deploying deep learning-based inspection systems on production lines. |
|---|---|
| Autores principais: | Carvalho, Miguel Martins |
| Assunto: | Foreign Object Detection Computer Vision YOLO Food Safety Deep Learning Synthetic Dataset |
| Ano: | 2026 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | The presence of foreign objects in food processing lines poses significant risks to consumer safety, product quality, and regulatory compliance. Traditional inspection methods such as manual visual checks or metal detectors are limited in their ability to detect non-metallic con- taminants like plastic, wood, or glass, especially when mixed with vegetable products. In re- sponse to this challenge, this work proposes the development of a real-time, computer vision system capable of detecting foreign materials on conveyor belts in a frozen vegetable factory. The proposed approach relies on RGB image acquisition and deep learning models, particu- larly object detection architectures such as YOLO (You Only Look Once), enhanced with at- tention mechanisms to improve small object recall. This document includes an overview of classical and modern vision techniques, the selection of suitable tools and models, and a defined work plan for implementation, calibration, testing and results. The expected outcome is a lightweight, accurate detection model integrated into a prototype system capable of operating in real-time under industrial conditions. This work aims to contribute to food safety automation in compliance with EU standards and to demon- strate the viability of deploying deep learning-based inspection systems on production lines. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.