Implementación de un prototipo de sistema de alerta para conductores distraídos y somnolientos de vehículos basado en visión artificial.

Abstract

In this titling work, a prototype warning system for vehicle drivers was implemented that detects and alerts about the presence of distraction and drowsiness using artificial vision. The system consists of four stages: qualifier training for the cell phone, image acquisition, processing, detection, and alarm activation. In the first stage, the preparation of the classifier was performed using the Haartraining technique to obtain unique characteristics of the object under study; in the second stage, a Raspberry Pi 3 B processor was used, which receives the real-time video recorded by the camera. The third stage is responsible for processing the video looking for characteristics of the state of the eyes, head settlement and cell phone; Once the processing has been carried out, the detection of drowsiness and distraction is carried out so that the alert stage will then start operating. The objective is selected by fixing the eyes, in which it identifies their state (open or closed), head settlement, and detection of cell phone use. Once these points are detected, a signal will be sent to the ports of the Raspberry, where an alert message will be emitted. The system algorithm was developed in Python for its diversity of libraries. Besides, the OPEN CV library was used for the variety of commands for image processing. As a result of the tests it was determined that the optimal distance is 60 cm and there should be not objects in the imaginary line between the camera and the driver's face. The speed that allows the system to process and send the information is 3.38 miliseconds and the system works with lighting greater than 2000 lux. It is recommended to extend the testing time of the prototype with the proposal to identify improvements to its design.