4D radar preprocessing balances accuracy and speed for driving safety
Accuracy- and Real-Time-Aware 4D Radar Preprocessing for Autonomous Driving Perception Systems
Computer Vision and Pattern Recognition
Summary
Autonomous cars need sensors that work well even in bad weather. The authors show how to prepare 4D radar data so that it keeps important details about objects, removes noisy points, and can run fast on limited hardware. They also created a way to check if these methods work well for real-time use in cars, balancing accuracy and speed. This helps make self-driving cars detect things more reliably in all conditions.
What this means in practice
- •For autonomous vehicle engineers: Improve perception modules by preprocessing 4D radar data to retain key object shapes and reduce noise without overloading limited onboard hardware.
- •For embedded systems developers: Evaluate and select radar preprocessing algorithms using Embedded & NetScore to balance accuracy, speed, and model complexity for resource-constrained devices.
Authors
Woo-Jin Jung, Dong-Hee Paek, Jeong-Su Park, Seung-Hyun Kong
Abstract
4D radar has emerged as a promising next-generation sensor for improving the robustness of autonomous driving perception systems because of its stable sensing capability under adverse weather conditions. However, deploying 4D radar in embedded environments with limited hardware resources requires radar-representation preprocessing that jointly considers perception accuracy, real-time performance, and computational complexity. This paper proposes a preprocessing framework for 4D-radar-based 3D object detection. First, Percentile-based 3D Shape Preservation (P3DP) extracts point clouds from radar tensors while preserving object-shape information and suppressing noise and false alarms. Second, Multi-frame-based Noise Point Discrimination using Kernel Density Estimation (MF-KDE) improves the density and reliability of sparse radar point clouds. Finally, Embedded \& NetScore (ENS) evaluates suitability for embedded deployment by jointly considering accuracy, real-time performance, adverse-weather robustness, and model complexity.