LCF3D: A robust and real-time late-cascade fusion framework for 3D object detection in autonomous driving

2026-01-01

Pattern Recognition

AI Summary

This work addresses the challenge of generalizing 3D object detection in autonomous driving when sensor configurations differ between training and testing, particularly involving RGB images and LiDAR point clouds. The authors propose LCF3D, a novel framework that integrates late fusion with cascaded fusion strategies: it leverages 2D–3D detection matching and late fusion to suppress LiDAR false positives, while simultaneously generating 3D frustum proposals from unmatched RGB detections to recover missed objects. Evaluated on the KITTI and nuScenes datasets, LCF3D significantly outperforms LiDAR-only methods, demonstrating superior performance especially on challenging classes such as pedestrians, cyclists, and motorcycles. The approach achieves robust, real-time, and cross-domain generalizable multimodal 3D object detection.

Technology Category

• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Problem

• 3D object detection
• sensor fusion
• autonomous driving
• LiDAR
• RGB camera

Innovation

• late fusion
• cascade fusion
• multimodal fusion
• 3D object detection
• domain generalization