Recognition of Brazilian vertical traffic signs and lights from a car using Single Shot Multi box Detector
This document presents a system for recognizing Brazilian traffic signs and lights using artificial intelligence. The main objective of the system is to contribute to road safety by alerting drivers to potential risks such as speeding, alcohol consumption, and cell phone use, which could lead to sev...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Estado: | Versión publicada |
| Fecha de publicación: | 2023 |
| País: | Brasil |
| Institución: | Universidade Federal de Uberlândia (UFU) |
| Repositorio: | Repositório Institucional da UFU |
| Idioma: | inglés |
| OAI Identifier: | oai:repositorio.ufu.br:123456789/39298 |
| Acceso en línea: | https://repositorio.ufu.br/handle/123456789/39298 https://doi.org/10.14393/ufu.di.2023.525 |
| Access Level: | acceso abierto |
| Palabra clave: | Artificial intelligence MobileNet SSD Traffic signs Inteligência artificial Sinais de trânsito CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAO::PROCESSAMENTO GRAFICO (GRAPHICS) Computação Estruturas de dados (Computação) Trânsito - Controle eletrônico |
| Sumario: | This document presents a system for recognizing Brazilian traffic signs and lights using artificial intelligence. The main objective of the system is to contribute to road safety by alerting drivers to potential risks such as speeding, alcohol consumption, and cell phone use, which could lead to severe accidents and jeopardize lives. The system’s core contribution lies in its ability to accurately detect and classify various traffic signs and lights, providing crucial warnings to drivers to prevent potential hazards. To achieve this, the system used the light version of the Single Shot Multibox Detector called SSD-Lite using Mobilenet version 2 and Mobilenet version 3 as base networks for feature extraction. The optimal Mobilenet version was selected based on performance evaluations to ensure a Mean Average Precision (mAP) higher than 80%, which guarantees reliable detection results. The dataset used for training and evaluation comprises images extracted from YouTube traffic videos, each meticulously annotated to create the necessary labels for model training. Through this extensive experimentation, the system demonstrates its efficacy in achieving accurate and efficient traffic sign and light detection. The results of the experiments are compared with other existing approaches that focus on detecting only one type of traffic sign or employ different network types. The proposed system outperforms these comparative works, showcasing its superiority in handling various traffic sign and light classes by providing a dedicated dataset for Brazilian traffic sign and light |
|---|