KPRC: Kernel-Point Convolutional Radar Codec for 4-D Millimeter-Wave Point Cloud Compression
IEEE Communications Letters · 2026 · 30 · 2198-2202
DOI: 10.1109/lcomm.2026.3702033 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Communications Letters, 2026. Verfasst von Mengjiang Sun und 3 weiteren Autor:innen. Die Arbeit nennt 13 Literaturangaben. Gefördert von: Natural Science Foundation for Excellent Young Scholars of Jiangsu Province, Open Fund of National Key Laboratory on Electromagnetic Environmental Effects and Electro-optical Engineering, National Natural Science Foundation of China.
Abstract (Originalsprache)
With the growing deployment of millimeter-wave radars in vehicle-to-everything (V2X) networks for cooperative perception, exchanging radar sensing data among participants can improve overall situational awareness. Lossless sharing of full-resolution radar data places a heavy burden on communication resources, whereas lossy compression can exploit the sparsity structure of radar point clouds to reduce bitrate. In this letter, we propose an end-to-end kernel-point radar codec (KPRC) for 4-dimensional (4-D) millimeter-wave radar point cloud compression and reconstruction. To match the characteristics of radar point clouds, KPRC uses a normalized radar kernel domain in which radial velocity participates in neighbor search and kernel correlation, while SNR is reconstructed as an attribute. Supervised training is performed on point clouds extracted from millimeter-wave radar measurements collected and processed in a roadside traffic scenario. The encoder maps each frame to a quantized latent vector for transmission, and the decoder reconstructs the radar point cloud from the received latent. Compared with the geometry-based point cloud compression (G-PCC) method, KPRC achieves lower rates at comparable reconstruction accuracy in our experiments.
Förderung
- Natural Science Foundation for Excellent Young Scholars of Jiangsu Province
- Open Fund of National Key Laboratory on Electromagnetic Environmental Effects and Electro-optical Engineering
- National Natural Science Foundation of China
- SEU Innovation Capability Enhancement Plan for the Doctoral Students Grant CXJH_SEU
Zitieren
APA
Sun, M., Chen, P., Li, Z., & Cao, Z. (2026). KPRC: Kernel-Point Convolutional Radar Codec for 4-D Millimeter-Wave Point Cloud Compression. IEEE Communications Letters, 30, 2198-2202. https://doi.org/10.1109/lcomm.2026.3702033
Vancouver
Sun M, Chen P, Li Z, Cao Z. KPRC: Kernel-Point Convolutional Radar Codec for 4-D Millimeter-Wave Point Cloud Compression. IEEE Communications Letters. 2026;30:2198-2202. doi:10.1109/lcomm.2026.3702033
BibTeX
@article{Sun2026,
author = {Sun, Mengjiang and Chen, Peng and Li, Zihang and Cao, Zhenxin},
title = {KPRC: Kernel-Point Convolutional Radar Codec for 4-D Millimeter-Wave Point Cloud Compression},
journal = {IEEE Communications Letters},
year = {2026},
volume = {30},
pages = {2198--2202},
doi = {10.1109/lcomm.2026.3702033}
}
EUVIMED