Publicação
Combination of physiological signals and image processing to detect driver drowsiness and distraction
| Resumo: | Over 1.3 million individuals lose their lives in road accidents each year, leaving behind broken families and communities. Road safety has become a global concern, with significant efforts directed toward preventing accidents and improving transportation systems. This thesis presents a comprehensive exploration of the current driver-state monitoring systems, aiming to enhance road safety. The study delves into the creation of a multimodal driver monitoring system, focusing on the user’s heart rate and image processing techniques. Our goal with this hybrid approach is to develop a system that is cost-effective and unobtrusive, that suits a wide range of vehicles. The work developed under this thesis involves the integration of heart rate data from a consumer-grade wearable for driver monitoring. Additionally, it employs machine learning models to process collected images by an in-vehicle camera, with the ultimate goal of detecting drowsiness or distraction. This process is done by extracting the region of interest of each collected frame and then using it as input to a model that classifies the driver state, into normal, drowsy, or distracted. The combination of the physiological data and the image processing results can then be used to trigger vibratory alerts to the driver, that are sent through the wearable device. In order to assess the reliability of the system, experimental procedures were used. The best-performing model showed decent accuracy, and the face detection algorithm achieved a high detection rate. This image processing module can be implemented in a real-time system, however, the complete prototype that was developed in Python does not operate at the required speed for a real vehicle. Ultimately, this work endeavors to contribute to reducing road accidents and enhancing driver security. It aspires to provide an effective approach to driver state monitoring, with potential applications in various contexts, from individual vehicles to commercial fleets. |
|---|---|
| Autores principais: | Oliveira, Daniel Sousa |
| Assunto: | Drowsiness Security system Image processing Driver monitoring Physiological signals Distraction detection |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Aveiro |
| Idioma: | inglês |
| Origem: | RIA - Repositório Institucional da Universidade de Aveiro |
| Resumo: | Over 1.3 million individuals lose their lives in road accidents each year, leaving behind broken families and communities. Road safety has become a global concern, with significant efforts directed toward preventing accidents and improving transportation systems. This thesis presents a comprehensive exploration of the current driver-state monitoring systems, aiming to enhance road safety. The study delves into the creation of a multimodal driver monitoring system, focusing on the user’s heart rate and image processing techniques. Our goal with this hybrid approach is to develop a system that is cost-effective and unobtrusive, that suits a wide range of vehicles. The work developed under this thesis involves the integration of heart rate data from a consumer-grade wearable for driver monitoring. Additionally, it employs machine learning models to process collected images by an in-vehicle camera, with the ultimate goal of detecting drowsiness or distraction. This process is done by extracting the region of interest of each collected frame and then using it as input to a model that classifies the driver state, into normal, drowsy, or distracted. The combination of the physiological data and the image processing results can then be used to trigger vibratory alerts to the driver, that are sent through the wearable device. In order to assess the reliability of the system, experimental procedures were used. The best-performing model showed decent accuracy, and the face detection algorithm achieved a high detection rate. This image processing module can be implemented in a real-time system, however, the complete prototype that was developed in Python does not operate at the required speed for a real vehicle. Ultimately, this work endeavors to contribute to reducing road accidents and enhancing driver security. It aspires to provide an effective approach to driver state monitoring, with potential applications in various contexts, from individual vehicles to commercial fleets. |
|---|