Bayesian curved lane estimation for autonomous driving

Abstract

Several pieces of research during the last decade in intelligent perception are focused on the development of algorithms allowing vehicles to move efficiently in complex environments. Most of existing approaches suffer from either processing time which do not meet real-time requirements, or inefficient in real complex environment, which also doesn't meet the full availability constraint of such a critical function. To improve the existing solutions, an algorithm based on curved lane detection by using a Bayesian framework for the estimation of multi-hyperbola parameters is proposed to detect curved lane under challenging conditions. The general idea is to divide a captured image into several parts. The trajectory is modeled by a hyperbola over each part, whose parameters are estimated using the proposed hierarchical Bayesian model. Compared to the existing works in the state of the art, experimental results prove that our approach is more efficient and more precise in road marking detection. Keywords Autonomous driving • embedded camera • road marking • multi-hyperbola • Bayesian framework

Vision-Based Autonomous Vehicle Systems Based on Deep Learning: A Systematic Literature Review

“…Following that, a massive contribution to lane and curve detection along with tracking was presented by studies where 95.5% road scene extraction was demonstrated, for example, in [79], for lane edge segmentation without manual labelling using a modified CNN architecture. As discussed in previous sections, challenges such as higher computational cost [81], insufficient for far field of view [82], not testing in complex scenarios [79] and poor luminance made some proposals tough for practical implementation in present AVS.…”

“…The Bayesian method for estimating multihyperbola parameters splitting frames in multiple patches was demonstrated by Fakhfakh et al to recognize curved lanes under difficult conditions using [82]. The lane line was represented on each component by a hyperbola which was determined using the proposed Bayesian hierarchical model with an average of 91.83% true positive rate (TPR) on the ROMA dataset.…”

Lane Detection in Autonomous Vehicles: A Systematic Review

“…For example, machine learning and deep learning are the AI algorithms that detect lanes. Unfortunately, most traditional lane detection systems suffer from either processing time that does not meet real-time needs or inefficiency in a complex environment that also fails to meet the total availability restriction of such a core function [45]. The two branches of AI-based methodology described in this paper are machine learning and deep learning-based techniques.…”

Non-smooth Bayesian learning for artificial neural networks

“…In this case, MCMC techniques are generally used to sample coefficients from the target posterior (Fakhfakh et al. 2020b ). The main limitation of such techniques lies in the high complexity level, especially when multidimensional data are handled.…”

“…However, the main difficulty is to derive analytical closed-form expressions of the estimators because of the posterior distribution form which can be complex if sophisticated priors are used, such as those promoting sparsity. In this case, MCMC techniques are generally used to sample coefficients from the target posterior Fakhfakh et al (2020b). The main limitation of such techniques lies in the high complexity level, especially when multidimensional data are handled.…”