Sistema embarcado para o monitoramento da sonolência e assistência ao condutor

• Ano de defesa: 2018
• Autor(a) principal: Soares, Gabriel Cambraia
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Lavras; Programa de Pós-Graduação em Engenharia de Sistemas e Automação; UFLA; brasil; Departamento de Engenharia
• Palavras-chave em Português: Sono – Identificação biométrica – Algorítmos computacionais; Olhos – Movimentos – Medição; Sistemas avançados de assistência ao motorista; Sleep – Biometric identification – Computer algorithms; Eye – Movements – Measurement; Advanced driver assistance systems; Sistemas de Computação
• Link de acesso: http://repositorio.ufla.br/jspui/handle/1/28993

With the advancement of technology and the high number of accidents involving land vehicles caused by drowsiness, assistance systems have been developed to monitor characteristics of drivers that indicate drowsiness. At the same time, the rise of technology related to smartphones and their spread in the different socioeconomic layers is a good opportunity for the diffusion of assistance systems, since most of the drowsiness identification modules are shipped in modern vehicles. Thus, this work proposes the development of a smartphone application for the monitoring of driver's drowsiness and sleepiness, based on the analysis of the state of their eyes according to PERCLOS indicator. The identification of the eyes is done by Local Binary Pattern classifiers, with the advantage of being fast and of low computational complexity, a very important requirement in mobile applications. These were trained with specific data sets and tested by four individuals, diversified in age, gender, and tonality of the eyes, in an environment with natural light. The occurrence of false positives was null and the average accuracy of open and closed classifiers was 84.16% and 73.23%, respectively. The identification of drowsiness was evaluated in an experiment composed by vigilant drivers, who had been awake for two hours and drowsy, who had been awake for eighteen hours. The PERCLOS of drowsy people presented an average value of 209.78% higher than that of vigilantes, and the symptom of sluggishness of the eyes during the blink of drowsy people was clearly noticed. The proposed sleepiness detection system operates at short intervals and alerts on events where closed eyes correspond to at least 90% of the total eyes found. Good results were obtained in the identification of the sleepiness, since in the ten simulations practiced the detection was committed, with an average time of 2.034 seconds. In the future, the application will be improved with a better user interface and integration of other information, such as vehicle speed, to maintain an efficient interaction with the driver.