Multi-Modal Fusion of Panoramic Reflectivity Images and Point Clouds from LiDAR for 3D Road Object Detection = 3차원 도로 객체 감지를 위한 LiDAR의 파노라마 반사율 이미지와 포인트 클라우드의 다중 모달 융합

목차 (Table of Contents)

• 1 INTRODUCTION 1
• 1.1 Overview 1
• 2 BACKGROUND 7
• 2.1 LiDAR 7
• 2.1.1 Overview of LiDAR 7
• 2.1.2 Types of LiDAR Sensors 9
• 2.2 LiDAR Data Representation 11
• 2.2.1 Raw Data 11
• 2.2.2 Point Cloud 15
• 2.2.3 Panoramic Attribute Image 16
• 2.3 Object Detection 19
• 2.3.1 2D Object Detection 19
• 2.3.2 3D Object Detection 20
• 2.4 LiDAR-based Sensor Fusion 22
• 3 PROPOSED 3D ROAD OBJECT DETECTION METHOD 24
• 3.1 Problem Statement 24
• 3.2 Overall Architecture of PIP-fusion 3D 26
• 3.3 Data Preparation 29
• 3.4 Cross-dimension Mapping 34
• 3.5 Clustering and 3D Bounding Box Estimation 38
• 3.6 IoU-based Filtering 42
• 4 EXPERIMENTAL 47
• 4.1 Datasets and Setup 47
• 4.2 2D Object Detector Analysis 51
• 4.3 3D Object Detector Analysis 54
• 4.4 Performance Comparison with LiDAR-only Baseline 56
• 4.5 Performance Comparison with LiDAR–Camera Fusion 58
• 5 CONCLUSION 64
• REFERENCES 66