Publications

Detailed Information

Optimization of ROI-based LiDAR sampling in on-road environment for autonomous driving

DC Field Value Language
dc.contributor.advisorHyuk-Jae Lee-
dc.contributor.authorPham Quan Dung-
dc.date.accessioned2020-10-13T02:53:44Z-
dc.date.available2020-10-13T02:53:44Z-
dc.date.issued2020-
dc.identifier.other000000161338-
dc.identifier.urihttps://hdl.handle.net/10371/169302-
dc.identifier.urihttp://dcollection.snu.ac.kr/common/orgView/000000161338ko_KR
dc.description학위논문 (석사) -- 서울대학교 대학원 : 공과대학 전기·정보공학부, 2020. 8. Hyuk-Jae Lee.-
dc.description.abstractLight detection and ranging (LiDAR) 센서는 최근 로보틱스와 자율 주행을 비롯한 여러 분야에서 사용되고 있다. 이런 LiDAR 센서는 다른 센서보다 낮은 해상도가 특징으로, 효과적인 샘플링 알고리즘을 설계하는 것이 필수적이다. 자율 주행에 적용되는 LiDAR 샘플링 알고리즘의 경우 도로의 복잡한 환경에서도 강인하게 높은 품질로 reconstruction을 하는 것이 목표이다. 이를 위해 현행 ROI 기반 샘플링 알고리즘은 시멘틱 정보를 이용하고 있다. 하지만, 객체, 도로, 배경 등에 따른 sampling rate는 지금까지 충분히 논의되지 않았고, 이로 인해 종합적인 reconstruction 품질이 저하될 수 있다. 이러한 문제를 해결하기 위해, 본 논문에서는 객체, 도로, 배경에 따른 sampling budget ratio를 도출할 수 있는 방법을 제안한다. 이 방법은 객체, 도로, 배경의 특성이 샘플링 이전에 선행 지식으로 주어져 있다는 가정을 이용한다. 제안하는 sampling budget을 적용한 결과, 현행 알고리즘보다 객체에 대한 mean-absolute-error (MAE)는 최대 45.92% 감소하였을 뿐만 아니라 전반적인 MAE 또한 3.36% 감소하였고, 도로에 대한 MAE는 오직 54.18% 감소하였다.-
dc.description.abstractIn recent years, light detection and ranging (LiDAR) sensors have been applied in several situations, including robotics and autonomous driving. However, LiDAR sensors have relatively low resolutions. Therefore, it is imperative to design an effective sampling algorithm for LiDAR sensors. To manage complex on-road environments, conventional ROI-based LiDAR sampling algorithm utilizes semantic information to achieve robust and high reconstruction quality. However, the ratio between sampling rates of objects, roads, and background areas is not thoroughly investigated. Therefore, the overall reconstruction quality may be degraded. To address this problem, this study presents a proposed method to examine the sampling budget ratio between objects, roads, and background areas, under the assumption that characteristics of objects, roads, and background areas are known prior to sampling. Experimental results depict a significant reduction in the mean-absolute-error (MAE) of the object region, road region and overall region by up to 45.92%, 54.18% and 3.36% under the proposed method, respectively, compared to the conventional method.-
dc.description.tableofcontentsChapter 1. Introduction 1
1.1. Overview 1
1.2. Light detection and ranging sensor LiDAR sampling 1
Chapter 2. Background 4
2.1. Definition of a sampling problem 4
2.2. Oracle Random Sampling 4
2.2.1 Sampling Model 4
2.2.2 Oracle Random Scheme 5
2.3. ROI-based LiDAR sampling algorithm 6
Chapter 3. Proposed method 8
3.1. Analytical method 8
Chapter 4. Experimental results 15
4.1. Dataset 15
4.2 Quantitative evaluation 16
Chapter 5. Conclusion 20
Appendix 21
References 31
초 록(Abstract in Korean) 32
-
dc.language.isoeng-
dc.publisher서울대학교 대학원-
dc.subjectSampling Algorithm-
dc.subjectLight Detection and Ranging Sensor-
dc.subjectLiDAR-
dc.subjectAutonomous Driving-
dc.subjectOn-road Environment-
dc.subjectROI-based Sampling-
dc.subject자율 주행-
dc.subject도로 환경-
dc.subjectROI 기반 샘플링-
dc.subject.ddc621.3-
dc.titleOptimization of ROI-based LiDAR sampling in on-road environment for autonomous driving-
dc.typeThesis-
dc.typeDissertation-
dc.contributor.AlternativeAuthor팜 콴 덩-
dc.contributor.department공과대학 전기·정보공학부-
dc.description.degreeMaster-
dc.date.awarded2020-08-
dc.identifier.uciI804:11032-000000161338-
dc.identifier.holdings000000000043▲000000000048▲000000161338▲-
Appears in Collections:
Files in This Item:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share