A Lane Detection Method Based on a Ridge Detector and Regional G-RANSAC

Area of Science:

• Computer Vision
• Robotics
• Autonomous Systems

Background:

• Lane detection is critical for autonomous vehicle safety.
• Existing methods struggle with adverse conditions like abnormal illumination and lane occlusion.

Purpose of the Study:

• To propose a novel and robust lane-division-lines detection method.
• To enhance the safety and reliability of autopilot systems.

Main Methods:

• Image conversion to aerial view to leverage lane geometry.
• Feature point extraction using a ridge detector and adaptable neural network (ANN).
• Lane line fitting with an improved random sample consensus (RANSAC) algorithm, termed Gaussian distribution RANSAC (G-RANSAC).

Main Results:

• The proposed method achieves high true-positive rates (TPR) across various scenarios (up to 99.02%).
• A new metric, the lane departure index (LDI), was introduced to quantify lane departure accuracy.
• Superior performance demonstrated compared to existing methods in diverse testing conditions.

Conclusions:

• The novel lane detection method offers significant improvements in accuracy and robustness.
• The G-RANSAC and ANN integration provides effective lane feature extraction and noise reduction.
• This method contributes to safer and more reliable autonomous driving systems.