Deep Learning for Autonomous Vehicle Traffic Sign Interpretation and Compliance

Abstract

Consequently, in this paper, a novel and cognitive deep-learning approach and edge artificial intelligence tool is introduced in order to robustly interpret traffic signs to facilitate intelligent cars to be able to be tested 0n the fly in any single country, and outside road infrastructure testing and to generalize the traffic sign compliance interpretations with all other international regulations, by a unified deep-learning based digital solid infrastructure for any countries. The proposed cognitive and accurate interpretation of traffic signs insured that the car can assure its full safety testing of all functions in each country. SOS’ traffic assistance adopted the proposal as an accuracy model for any region casewith cultural versatility. Deep learning (DL) or edge artificial intelligence is considered a superior cognitive platform for the accurate interpretation of a traffic sign by a unified international language in any region. Slight region adaptation of the car parameters and traffic signs provide accurate road traffic sign compliance monitoring in any region. A huge dataset from the German regional traffic sign data and the Russian regional traffic sign defines a unified language for the traffic sign test cycle after autonomous car performance repairs in every region by a high cognitive level of smart surveillance. SOS (Safe-Our-Selves) is a project acronym promoted by Yandex Company to test internationally all their driving functions. SOS competence is the capability of the system to guarantee the performance of driving functions on the test bench and on the road regardless of origin of the traffic signs. SOS competence is an important competitive and safety requirement for the SOPov or IV SOPov. After the theory part, to communicate the vision of this traffic sign visualization, an image example is fused by 10 interpretation of German traffic signs, Russian traffic signs and international traffic signs which has been detected, two by two separately by implementing the proposed artificial intelligent deep-learning tool and intelligent enhancement.