A Robust Lane Detection Model Using Vertical Spatial Features and Contextual Driving Information

Area of Science:

• Computer Vision
• Robotics
• Artificial Intelligence

Background:

• Lane detection is critical for autonomous vehicle navigation.
• Existing models struggle with complex driving scenes, including poor lighting and occlusions.
• Unclear and occluded lane lines pose significant challenges for current lane detection algorithms.

Purpose of the Study:

• To develop a robust lane detection model for complex driving environments.
• To enhance the detection of unclear and occluded lane lines.
• To improve the reliability of lane detection in unmanned vehicles.

Main Methods:

• Proposed a novel lane detection model incorporating vertical spatial features and contextual driving information.
• Introduced a feature merging block to enhance contextual information for the network.
• Developed an information exchange block combining spatial and dilated convolutions for improved pixel-level information transfer.

Main Results:

• The proposed model demonstrates superior robustness and precision in detecting lane lines compared to state-of-the-art methods.
• Experimental results validate the model's effectiveness across diverse complex driving scenarios.
• The model successfully addresses challenges posed by unclear and occluded lane lines.

Conclusions:

• The novel lane detection model offers enhanced performance in complex driving conditions.
• The integration of vertical spatial features and contextual information significantly improves lane detection accuracy.
• This research contributes to safer and more reliable autonomous driving systems.