Many lane detection methods only consider single frame information and often ignore the contextual information between consecutive frames, which is not robust enough in the absence of lane visual information. In practice, lane detection is usually processed in a dynamic environment. If the semantic relationship between multiple frames is learned, the complementary information from adjacent frames can be used to make up for the absence of visual information in some certain frames, and thus the accuracy of lane detection will be improved. Based on this idea, we propose a convolutional GRU (ConvGRU) model to fuse continuous multi-frame lane feature information and enhance the semantic information of the current frame as well. Moreover, for the current lane dataset lacks datasets in complex scenarios, we generate four more challenging lane scene datasets in the original TuSimple dataset through the style transfer algorithm to verify the robustness of the model. In different complex lane scenes, our method can achieve the state-of-the-art performance in terms of accuracy, precision and F1-Measure. Our code is available at https://github.com/Cuibaby/ConvGRULane.
Enhancing Robustness of Lane Detection Through Dynamic Smoothness
Lect. Notes Electrical Eng.
International Conference on Autonomous Unmanned Systems ; 2021 ; Changsha, China September 24, 2021 - September 26, 2021
Proceedings of 2021 International Conference on Autonomous Unmanned Systems (ICAUS 2021) ; Kapitel : 16 ; 148-161
2022-03-18
14 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Robustness of Lane Mark Detection with Wide Dynamic Range Vision Sensor
British Library Conference Proceedings | 1996
|Receiving device for track smoothness detection and track smoothness detection system
Europäisches Patentamt | 2022
|Gradient-Enhancing Conversion for Illumination-Robust Lane Detection
Online Contents | 2013
|