A preliminary study on updating a detected traffic cone's coordinates obtained from a stereo camera on a pseudo high definition map database

Miguel Luis R. Lagahit, Yi Hsing Tseng

Research output: Contribution to conferencePaperpeer-review

Abstract

Research and development on Autonomous Vehicles (AVs) or self-driving cars have been escalating around the world these past few years. One of those researches is about HD (High-Definition) Maps, which are basically very detailed 3D (three-dimensional) maps produced from mobile mapping techniques that uses a combination of GPS, IMU, LiDAR, cameras, and etc., as its sensors. HD Maps helps the AV better position itself and other objects (traffic cones, traffic signs, lane markings, etc.) on the road by providing all of the necessary geometric and semantic information. This research will propose a preliminary method on how to update said HD Maps. The methodology will mainly consist of: (1) running a pre-trained YOLOv3, an object detection system, on a calibrated stereo camera to detect the traffic cone, (2) applying photogrammetric techniques to determine its 3D position, (3) uploading and updating the observed coordinates to the pseudo HD Map database, which is hosted on a local PostgreSQL platform, (4) executing all of it at the same time on a Python based program, and then (5) visualizing it in a simulated scenario through a browser based application using Cruise Automation's Worldview. Results have shown centimeter level accuracy in terms of obtained distance of the detected traffic cone from the camera setup, and a successful upload and or update of the traffic cone's coordinates to a database. In future works, additional sensors such as LiDAR may be included in the detecting and positioning framework, additional computer vision or image processing techniques such as feature tracking may also be included, and objects of interest can be expanded to include other traffic management objects such as traffic signs and other lane markings.

Original languageEnglish
Publication statusPublished - 2020 Jan 1
Event40th Asian Conference on Remote Sensing: Progress of Remote Sensing Technology for Smart Future, ACRS 2019 - Daejeon, Korea, Republic of
Duration: 2019 Oct 142019 Oct 18

Conference

Conference40th Asian Conference on Remote Sensing: Progress of Remote Sensing Technology for Smart Future, ACRS 2019
CountryKorea, Republic of
CityDaejeon
Period19-10-1419-10-18

All Science Journal Classification (ASJC) codes

  • Information Systems

Fingerprint Dive into the research topics of 'A preliminary study on updating a detected traffic cone's coordinates obtained from a stereo camera on a pseudo high definition map database'. Together they form a unique fingerprint.

Cite this