Deep Learning Techniques for Road Lane Detection for Traffic Control

Authors

  • T.Srilalith, Dr. N. Venkatesh

Keywords:

Autonomous driving, Driver assistance system, Semantic segmentation, Machine learning, Deep learning

Abstract

The development of intelligent traffic systems and autonomous driving has depended heavily on road lane identification in recent years. To improve road lane recognition accuracy and resilience, this research introduces a unique strategy that combines classic machine learning approaches with hybrid deep learning. The suggested approach combines the effectiveness of machine learning techniques for lane border classification and tracking with the advantages of convolutional neural networks (CNNs) for feature extraction and pattern identification. Using a hybrid model, the system retains the interpretability and flexibility of machine learning while gaining access to deep learning's high-dimensional data processing capabilities. This research presents a comparative analysis of many deep neural network designs that can infer surface normal information on the traditional KITTI road dataset under a range of demanding conditions. By testing it on three cutting-edge deep learning models—"Resnet-50," "Xception," and "MobileNet-V2"—we want to streamline the process of how current approaches interpret ground-related data and offer a solution. This will allow us to better understand and use the capabilities of these models. This comparative study's primary contribution has been to assess these networks' performance for edge deployment. Therefore, the little DNN model of MobileNet-V2 has been taken into consideration, which has about 80% less adjustable parameters than the other models. The acquired outcomes demonstrate that the suggested networks may get a segmentation accuracy of over about 96%, even in a variety of difficult situations. The method is tested on a large dataset with different surroundings and road conditions. Comparing experimental results to state-of-the-art techniques, one can observe notable gains in computing efficiency and detection accuracy. The integration of this sturdy lane identification technology into traffic management systems can facilitate the development of autonomous driving solutions that are more dependable and safer.

Downloads

Published

2024-07-29

How to Cite

T.Srilalith, Dr. N. Venkatesh. (2024). Deep Learning Techniques for Road Lane Detection for Traffic Control. Utilitas Mathematica, 121, 127–148. Retrieved from https://utilitasmathematica.com/index.php/Index/article/view/2011

Citation Check