Cross-Domain Generalization for LiDAR-Based 3D Object Detection in Infrastructure and Vehicle Environments

作     者：Zhi, Peng Jiang, Longhao Yang, Xiao Wang, Xingzheng Li, Hung-Wei Zhou, Qingguo Li, Kuan-Ching Ivanovic, Mirjana 

作者机构：Lanzhou Univ Sch Informat Sci & Engn Lanzhou 730000 Peoples R China Providence Univ Dept Comp Sci & Informat Engn Taichung 43301 Taiwan Univ Novi Sad Fac Sci Novi Sad 21000 Serbia 

出 版 物：《SENSORS》 (Sensors)

年 卷 期：2025年第25卷第3期

页      面：767-767页

核心收录：

学科分类：0710[理学-生物学] 071010[理学-生物化学与分子生物学] 0808[工学-电气工程] 07[理学] 0804[工学-仪器科学与技术] 0703[理学-化学] 

基　　金：National Key R&D Program of China [2023YFB4503903, 2020YFC0832500] National Natural Science Foundation of China [U22A20261, 61402210] Gansu Province Science and Technology Major Project-Industrial Project [22ZD6GA048, 23ZDGA006, 4E49EFF3] Gansu Province Key Research and Development Plan-Industrial Project [22YF7GA004] Gansu Provincial Science and Technology Major Special Innovation Consortium Project [21ZD3GA002] Fundamental Research Funds for the Central Universities [lzujbky-2024-jdzx15, lzujbky-2022-kb12, lzujbky-2021-sp43, lzujbky-2020-sp02, lzujbky-2019-kb51] Open Project of Gansu Provincial Key Laboratory of Intelligent Transportation [GJJ-ZH-2024-002] Science and Technology Plan of Qinghai Province [2020-GX-164] China Higher Education Institutions Industry-Academia-Research Innovation Fund for Digital Intelligence and Educational Projects [2023RY020] 

主　　题：3D object detection V2X cooperative perception LiDAR point clouds infrastructure sensors 

摘      要：In the intelligent transportation field, the Internet of Things (IoT) is commonly applied using 3D object detection as a crucial part of Vehicle-to-Everything (V2X) cooperative perception. However, challenges arise from discrepancies in sensor configurations between vehicles and infrastructure, leading to variations in the scale and heterogeneity of point clouds. To address the performance differences caused by the generalization problem of 3D object detection models with heterogeneous LiDAR point clouds, we propose the Dual-Channel Generalization Neural Network (DCGNN), which incorporates a novel data-level downsampling and calibration module along with a cross-perspective Squeeze-and-Excitation attention mechanism for improved feature fusion. Experimental results using the DAIR-V2X dataset indicate that DCGNN outperforms detectors trained on single datasets, demonstrating significant improvements over selected baseline models.