A Multi-Feature Fusion Approach for Road Surface Recognition Leveraging Millimeter-Wave Radar

Area of Science:

• Intelligent Transportation Systems
• Sensor Fusion
• Machine Learning for Autonomous Driving

Background:

• Accurate road surface recognition is vital for autonomous driving safety and comfort.
• Existing methods may lack robustness in diverse road conditions.

Purpose of the Study:

• To develop a multi-feature fusion approach for road surface identification using millimeter-wave radar.
• To enhance the perception capabilities of intelligent vehicles.

Main Methods:

• Utilized 24 GHz millimeter-wave radar for data acquisition.
• Extracted six-dimensional statistical features and wavelet transform features.
• Fused features into a 56-dimensional vector for classification.
• Employed Wide Neural Network, KNN, SVM, and Kernel methods as classifiers.

Main Results:

• Achieved a road surface type identification accuracy of 94.2% with 8865 real-world samples.
• Demonstrated the effectiveness of fusing statistical and wavelet features.
• Validated the method across 12 typical road surface types and conditions.

Conclusions:

• The proposed multi-feature fusion method provides an efficient and cost-effective road perception solution.
• Millimeter-wave radar shows significant potential for road environment sensing in autonomous driving.
• This research supports the advancement of autonomous driving technology through improved road surface identification.