Detección automática de carriles y vehículos en entornos urbanos utilizando técnicas de visión artificial

Abstract

Abstract: This thesis addresses the issue of road safety and the need for driver assistance systems that provide support in the face of risks associated with human error and adverse conditions in urban environments. As a starting point, preliminary research was conducted to establish a solid theoretical basis for selecting the tools and algorithms that yield the best results. The management and labeling of data sources were centralized using the Roboflow platform. This preparation was key to developing a functional prototype, which integrated two specialized models based on YOLO11 in its medium size: one dedicated to lane line detection and the other focused on vehicle detection. These components were grouped into a single architecture, achieving a unified system that operates in real time. Finally, through validation tests performed on the prototype, it was demonstrated that its ability to alert the driver meets the objectives of improving driving safety through computer vision technologies.