Improving the Accuracy of Lane Detection by Enhancing the Long-Range Dependence

Lane detection is a common task in computer vision that involves identifying the boundaries of lanes on road from an image or video. Improving accuracy lane great help to advanced driver assistance systems and autonomous driving cars identify keep correct lane. Current high-accuracy models are mainly based artificial neural networks. Among them, CLRNet latest famous model, which attains high accuracy. However, some scenarios, lower accuracy, we revealed this caused by insufficient global dependence information. In study, enhanced proposed new model called NonLocal (NLNet). algorithmic mechanism captures long-range dependence. NLNet employs acquire more information then applies acquired Feature Pyramid Network (FPN) for improving Using CULane dataset, trained NLNet. The experimental results showed outperformed state-of-the-art terms most particularly no-line scenario night scenario. This study very helpful developing accurate models.