Efficient and real-time lane detection using CUDA-based implementation

Abstract

Lane detection is an essential component of autonomous driving systems, enabling vehicles to accurately identify and follow road markings. In this paper, we look at an lane detection approach that integrates median filtering and the Hough transform. Median filtering is an essential pre-processing step for reducing noise and improving lane detection accuracy. However, given its high computational demands, optimization of this process is essential for real-time applications. To this end, we used CUDA for acceleration, taking advantage of its parallel computing capabilities to improve performance. We implemented and tested this optimised lane detection algorithm on the NVIDIA Jetson Nano and on a desktop, providing a comparative analysis of improvements in efficiency and speed. This approach highlights the potential of real-time path detection in embedded and high-performance computing environments.