Lane Detection Algorithm for Intelligent Vehicles in Complex Road Conditions and Dynamic Environments.

Area of Science:

• Computer Vision
• Robotics
• Artificial Intelligence

Background:

• Traditional lane detection methods suffer from low accuracy.
• Deep learning approaches often lack real-time performance for intelligent vehicles.
• Complex road conditions and dynamic environments pose challenges for existing algorithms.

Purpose of the Study:

• To develop a robust lane detection algorithm for intelligent vehicles.
• To overcome limitations of traditional and deep learning methods in accuracy and speed.
• To enhance lane detection capabilities in challenging driving scenarios.

Main Methods:

• Image preprocessing including distortion correction and edge detection using a superposition threshold algorithm.
• Inverse perspective transformation to obtain an aerial view of lanes.
• Lane line curve fitting using the random sample consensus (RANSAC) algorithm with a third-order B-spline model.
• Performance evaluation using road driving videos and the TuSimple dataset.

Main Results:

• Achieved 98.49% average detection accuracy and 21.5 ms processing time on road driving video.
• Achieved 98.42% average detection accuracy and 22.2 ms processing time on the TuSimple dataset.
• Demonstrated superior accuracy, real-time performance, and anti-interference ability compared to existing methods.

Conclusions:

• The proposed lane detection algorithm offers excellent accuracy and efficiency for intelligent vehicles.
• The algorithm significantly improves upon traditional and deep learning-based lane detection techniques.
• This advancement is crucial for enhancing intelligent vehicle driving assistance and safety.