멀티 스케일 적대적 생성 신경망을 이용한 라이다 채널 해상도 개선Lidar Super-resolution based on Multi-scale Generative Adversarial Network
- Other Titles
- Lidar Super-resolution based on Multi-scale Generative Adversarial Network
- Authors
- 김민정; 최승원; 허건수
- Issue Date
- Jun-2021
- Publisher
- 한국자동차공학회
- Citation
- 2021 한국자동차공학회 춘계학술대회, pp.488 - 492
- Journal Title
- 2021 한국자동차공학회 춘계학술대회
- Start Page
- 488
- End Page
- 492
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/191363
- Abstract
- To perform stable autonomous driving, it is essential to understand the dynamic environment through sensors.Sensors for object detection include cameras, lidars, and radar sensors. Especially, lidars can obtain relatively accurate threedimensional point cloud data by measuring the time reflected on objects after shooting laser pulses. In addition, as the intervalbetween pulses increases, the detection performance of distant objects decreases, high-channel lidar is necessary to determinethe shape of the distant object. However there is a problem that the more channels Lidar has, the more it costs. Therefore,research is actively underway to convert low-channel lidars into images and super-resolution them into high-channel lidars. Inthis paper, we present a method for converting low-channel lidar point clouds to high-channel lidar point clouds using multiscale adversarial generative neural networks (GANs). We project a 16-channel point cloud into a panoramic image, and thentrain the network to generate 64-channel point cloud by putting it as an input value into a neural network. The proposed algorithm has been verified using Carla simulation set.
- Files in This Item
-
Go to Link
- Appears in
Collections - 서울 공과대학 > 서울 미래자동차공학과 > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/191363)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.