Cornell University
Vision-based 3-D Pedestrian Tracking for Autonomous Vehicles Using Pseudo-LiDAR
Abstract
dc:description.abstractPreviously, a variety of pedestrian tracking methods for different sensing sys-tems have been studied. However, with limited methods of effectively com-paring the tracking performance using different systems, it is difficult to under-stand the performance gap of vision-only sensing system in pedestrian trackingunder the criteria of real-world applications. We propose a general multi-sensorpipeline that allows vison-only tracking to be directly compared with LiDAR-vision integrated tracking. The pipeline supports switching among differentsensing systems by transferring image-based depth information into 3-D pointcloud through Pseudo-LiDAR. Furthermore, by processing vision-only sensingdata through the proposed pipeline, we show that vision-only sensing systemachieves performance close to LiDAR for pedestrian tracking in 3-D space.
Degree
thesis:*- Name thesis:degree_name
- M.S., Mechanical Engineering
- Level thesis:degree_level
- Master of Science
- Discipline thesis:degree_discipline
- Mechanical Engineering
- Grantor
- Cornell University
- Year dc:date.issued
- 2020
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Liu, Ertai
- Chair dc:contributor.chair
-
- Campbell, Mark
- Committee member dc:contributor.committeemember
-
- Hariharan, Bharath
Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Bibid
- 13277728
- Dc Identifier Other
-
ProQuest Submission ID: 11006
ProQuest Publication ID: 28086909