Análise e fusão de imagens 2D e 3D com vistas para detecção e classificação de sinais de trânsito verticais em prol da segurança viária com veículos robóticos inteligentes
BRUNO, D. R. Analysis and merging of 2D and 3D images for detecting and classifying vertical traffic signs to the benefit of the road safety with intelligent robotic vehicles. 2020. 191 p. Tese (Doutorado em Ciências -Intelligent Robotic Vehicles are mainly applied for the benefit of the traffic accidents reduction, thus enabling to reduce human faults and recklessness with systems that use Computer Vision, Artificial Intelligence, Automation and other technologies to assist the driver in his
... the driver in his driving task. By applying robotization, it is also possible to increase the level of road safety by developing autonomous vehicles totally free of human control and which are programmed to navigate within traffic laws. Human failures in driving take responsability over 90% for fatal accidents worldwide. The main objective of this doctoral research was the study, proposal, development, adaptations and tests of a set of techniques and methods of Computer Vision and Artificial Intelligence, aiming at a 2D and 3D image fusion perception system, being more robust for detecting vertical road signs. A Fuzzy Visual Attention model was also developed, capable of analyzing the priority of each information detected through the perception system, thus enabling the decision making of the vehicle involving emergency situations (accidents and road works) to be supported. The Fuzzy Visual Attention model uses the priority values of each traffic sign as a basis. The Robotic Computer Vision system shall be capable of detecting, classifying and analyzing the priority of currently used vertical traffic signs that are functional for traffic involving human drivers in the real world and do not require signaling adaptations. The vision system should then assist a fully autonomous or semi-autonomous vehicle to navigate within local traffic rules, thus detecting important information such as maximum speed, mandatory stop, emergency cones and traffic light colors. In cases of autonomous navigation, only the system of perception and analysis of vertical traffic signs should be used. For semi-controlled navigation, that is, with the help of a human, the external vision system should work in conjunction with driver analysis and vehicle control data, activating automatic corrective routines based on errors detected in the task execution. Thus making it possible to avoid serious accidents related to disregarding traffic signs, due mainly to human errors and recklessness.