Design and implementation of an Advanced Driver Assistance System on a racing prototype in the marine environment using stereo vision and Convolutional Neural Networks

This thesis is based on the internship experience carried out at CVLab, a research laboratory of the University of Bologna that focuses on computer vision research topics, and UniBoAT, one of the motorsport teams of the University of Bologna that develops a prototype in the marine environment. Vehicles play a vital role in modern society. However, despite their importance, transportation poses significant negative externalities to human society, such as pollution, accidents, and human casualties. For example, road accidents are the leading cause of death among young people aged between 5 and 29 years, and they rank as the 8th leading cause of death across all age groups, surpassing diseases like HIV/AIDS and tuberculosis. In recent years, with the advancements in AI (Artificial Intelligence), devices known as ADAS (Advanced Driver-Assistance System) have been developed. ADAS refers to a group of electronic technologies that assist drivers in driving and parking functions. These systems utilize automated technology, such as sensors and cameras, to detect nearby obstacles or driver errors and respond accordingly. In the pursuit of a more precise driving system than one based on human capabilities, research on autonomous vehicles has shown promising results in reducing accidents. Additionally, autonomous vehicles have the potential to relieve individuals from the mental and physical burdens of long travels. One significant area of AI is computer vision, which enables computers and systems to derive meaningful information from digital images, videos, and other visual inputs. Object detection is a popular task in computer vision because it allows vehicles to detect lanes or perform person detection to improve safety. The major challenges in object detection involve classifying objects and determining their position, as images are subject to occlusions, deformation, and viewpoint variation.