SMURF: Spatial Multi-Representation Fusion for 3D Object Detection with 4D Imaging Radar

TL;DR

This paper introduces SMURF, a novel 3D object detection method using 4D radar that fuses multiple representations and density features to improve accuracy and real-time performance in adverse conditions.

Contribution

SMURF is the first to leverage multi-representation fusion with kernel density estimation for 4D radar-based 3D detection, enhancing robustness and efficiency.

Findings

Outperforms recent single-representation models on VoD and TJ4DRadSet datasets.

Achieves comparable results to radar-camera fusion methods using only radar data.

Maintains inference time under 0.05 seconds for most scans.

Abstract

The 4D Millimeter wave (mmWave) radar is a promising technology for vehicle sensing due to its cost-effectiveness and operability in adverse weather conditions. However, the adoption of this technology has been hindered by sparsity and noise issues in radar point cloud data. This paper introduces spatial multi-representation fusion (SMURF), a novel approach to 3D object detection using a single 4D imaging radar. SMURF leverages multiple representations of radar detection points, including pillarization and density features of a multi-dimensional Gaussian mixture distribution through kernel density estimation (KDE). KDE effectively mitigates measurement inaccuracy caused by limited angular resolution and multi-path propagation of radar signals. Additionally, KDE helps alleviate point cloud sparsity by capturing density features. Experimental evaluations on View-of-Delft (VoD) and TJ4DRadSet datasets demonstrate the effectiveness and generalization ability of SMURF, outperforming recently proposed 4D imaging radar-based single-representation models. Moreover, while using 4D imaging radar only, SMURF still achieves comparable performance to the state-of-the-art 4D imaging radar and camera fusion-based method, with an increase of 1.22% in the mean average precision on bird's-eye view of TJ4DRadSet dataset and 1.32% in the 3D mean average precision on the entire annotated area of VoD dataset. Our proposed method demonstrates impressive inference time and addresses the challenges of real-time detection, with the inference time no more than 0.05 seconds for most scans on both datasets. This research highlights the benefits of 4D mmWave radar and is a strong benchmark for subsequent works regarding 3D object detection with 4D imaging radar.