YOLO-Based Simultaneous Target Detection and Classification in Automotive FMCW Radar Systems.

Area of Science:

• Automotive Engineering
• Computer Vision
• Machine Learning

Background:

• Traditional automotive radar systems use separate stages for object detection and classification.
• This sequential approach can be inefficient and may miss crucial information.

Purpose of the Study:

• To develop a unified deep learning framework for simultaneous object detection and classification in automotive radar.
• To improve the efficiency and accuracy of automotive radar perception systems.

Main Methods:

• Utilized the You Only Look Once (YOLO) deep learning model.
• Applied the model to pre-processed automotive radar signals for integrated detection and classification.
• Validated the method using real-world automotive radar data.

Main Results:

• Achieved simultaneous detection and classification of targets with over 90% accuracy.
• Demonstrated superior performance compared to conventional methods like DBSCAN and Support Vector Machine (SVM).
• Showcased enhanced capabilities in detecting and classifying long-bodied vehicles.

Conclusions:

• The proposed YOLO-based method offers a more efficient and accurate approach to automotive radar perception.
• This unified detection and classification strategy significantly advances object recognition in autonomous driving systems.
• The method shows particular promise for identifying complex vehicle shapes in radar data.