Probabilistic risk assessment using major features of rural road curves via rear and front car cameras

Abstract

Rural and secondary roads inherently exhibit road curves and turns. This study aims to predict the risk associated with rural road curves exploiting major curve features such as curvature, slope type, and direction. The images acquired by rear and front cameras are utilized to capture the underlined features. Slope type is obtained by two-view images while direction and curvature are estimated by single-view front camera. Our approach is based on geometrical derivations using visual clues such as vanishing points and curb borders. Consequently, the impact of major features on the risk has been assessed using the Bayesian belief theory and network. The proposed model is expected to be an advanced driver assistant system for long distance drivers, which tackles with prominent risk components associated with road curves. In turn, this type of advanced driver assistance systems would become a critical part of autonomous vehicles