Modular Prototype of Artificial Vision for the Detection of Fatigue and Anti-drowsiness in Drivers of Land Vehicles

The purpose of this work was to build an artificial vision device that helps safeguard people’s lives from traffic accidents related to fatigue and anti-drowsiness, through an alarm that makes the driver react in time. This is an applied research with a quantitative approach. The iterative methodology was used as the basis of the research; in the first iteration, information was collected to define the requirements and associated risks. Subsequently, the system was built in four iterations, the first consisted of image acquisition and detection, the second the database was built in SQLite and the model was trained, the third consisted of sending alert messages, through the 4g module, and the fourth, referring to the activation of the different audible and visual peripherals to obtain the driver’s reaction. The Cascade Trainer GUI platform was used. Additionally, the model was trained with the same tool. Python was used for the face and eye detection programs, as well as the playsound library for the messages. The electronic design includes high-power LED lights, amplifier circuit for LED lights, GPRS or GSM module, Raspi (infrared camera), Raspberry PI 3B, speakers, power supplies. The design of the electronic detection was carried out in PCB Wizard and the design of the casing was developed in Freecad. The system was validated using the confusion matrix, from which indicators of accuracy, precision, specificity, and score above 90% were obtained, with which a device could be presented whose detection speed was determined in 1.2 s.