Adaptive Lane Detection for Indian Roads with faded markings

Abstract

Lane detection is a key component in Advanced Driver Assistance Systems (ADAS) and autonomous vehicles. However, detecting lanes on Indian roads is challenging due to faded lane markings, poor road maintenance, uneven lighting conditions, shadows, traffic congestion, and unstructured road environments. Traditional lane detection methods based on clear white markings often fail in such conditions. This project proposes an adaptive lane detection system specifically designed for Indian road scenarios. The system uses image processing techniques such as grayscale conversion, edge detection, region of interest selection, and Hough Transform along with machine learning or deep learning models like Convolutional Neural Networks (CNN) to detect lanes even when markings are partially visible or faded. This adaptive approach enhances road safety, reduces driver fatigue, and supports the development of intelligent transportation systems suitable for Indian traffic conditions.

Introduction

This project focuses on developing an adaptive lane detection system for Indian road conditions, where traditional lane detection methods often fail due to faded markings, poor maintenance, uneven lighting, heavy traffic, and unstructured road layouts. Lane detection is a key component of Advanced Driver Assistance Systems (ADAS) and autonomous driving, but most existing systems are designed for well-marked Western roads and perform poorly in real-world Indian environments.

Traditional image-processing-based methods rely heavily on clear lane markings and simple visual cues, achieving limited accuracy (around 70–85%) and becoming unreliable under shadows, night conditions, rain, fog, and lane occlusions. To overcome these limitations, modern approaches use Convolutional Neural Networks (CNNs) and deep learning, which can automatically learn robust features from raw images and improve detection in complex scenarios. However, many deep learning models still struggle due to lack of India-specific datasets and high computational requirements.

The proposed system introduces a hybrid approach combining image preprocessing and CNN-based deep learning. Preprocessing steps such as grayscale conversion, blurring, edge enhancement, and Region of Interest (ROI) extraction help remove noise and focus on the road area. The CNN model then performs feature extraction and semantic segmentation, classifying each pixel as lane or non-lane. The system is trained on a custom Indian road dataset covering day, night, curved roads, and adverse weather conditions.

Experimental results show strong performance, achieving approximately 95% accuracy in daytime, 90% at night, and reliable detection on curved roads, outperforming traditional methods and approaching state-of-the-art results. Overall, the system improves lane detection reliability in challenging Indian road environments, enhancing safety for ADAS applications through a more robust and adaptive deep learning-based approach.

Conclusion

The proposed Adaptive Lane Detection system effectively addresses the challenges of detecting lanes on Indian roads with faded or unclear markings. By integrating deep learning techniques such as CNN along with adaptive preprocessing methods, the system improves detection accuracy under varying lighting, shadows, and heavy traffic conditions. Unlike traditional methods that rely on fixed thresholds and clear markings, the proposed approach dynamically adapts to real-world scenarios. The model demonstrates improved robustness, reliability, and real-time performance. Overall, this system contributes to enhanced road safety and supports the development of intelligent transportation and ADAS technologies suitable for Indian environments.

References

[1] Z. Wang, W. Ren and Q. Qiu, Lane Net: Real-Time Lane Detection Networks for Autonomous Driving, arXiv preprint, Jul. 2018. Available: https://arxiv.org/abs/1807.01726 [2] R. M. Patel and V. Parekh, “Lane Detection in Autonomous Driving Using Enhanced Preprocessing and AI Techniques,” Int. Educ. and Res. J., vol. 11, no. 04, Apr. 2025. Available: https://ierj.in/journal/index.php/ierj/article/view/4 100 [3] H. S. Gowri Yaamini et al., “Lane and Traffic Sign Detection for Autonomous Vehicles: Addressing Challenges on Indian Road Conditions,” MethodsX, 2025. Available: https://pubmed.ncbi.nlm.nih.gov/39911905/ [4] A. Sai Hanuman and G. Prasanna Kumar, “Robust and real-time multi-lane and single lane detection in Indian highway scenarios,” E3S Web of Conferences, vol. 309, 01016, 2021. Available: https://doi.org/10.1051/e3sconf/202130 [5] J. Cao, C. Song, S. Song, F. Xiao and S. Peng, “Lane Detection Algorithm for Intelligent Vehicles in Complex Road Conditions and Dynamic Environments,” Sensors, vol. 19, no. 14, 2019. Available: https://doi.org/10.3390/s19143166 [6] K., Gkioxari, G., Dollar, P. & Girshick, R., “Mask R-CNN.” IEEE International Conference on Computer Vision (ICCV), 2017. Available: https://ieeexplore.ieee.org/document/8237584 [7] Kim, Z. and Kim, D. “Fast Lane Detection Based on Modified Hough Transform and Kalman Filter.” IEEE Transactions on Intelligent Transportation Systems, 2008. Available: https://ieeexplore.ieee.org/document/4598640

Copyright

Copyright © 2026 Gunja Ajay Kumar, Gorantla Manoj Kumar Reddy, Gopu Sasi Kumar Reddy, Gurram Nagesh, Dr. M. Baby Nirmala. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.