3D tracking of objects in real time
Hyttinen, Esa (2018)
2018
Automaatiotekniikka
Teknisten tieteiden tiedekunta - Faculty of Engineering Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2018-05-09
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201804261579
https://urn.fi/URN:NBN:fi:tty-201804261579
Tiivistelmä
This thesis aims to explore the problem of object tracking. This included reviewing existing applications and technologies related to the problem, and testing one approach via setting up a system that tracks obstacle location. Also, the suitability of the selected hardware was to be assessed. Obstacle tracking solutions could be used in various tasks with autonomous mobile machines, for example avoiding collisions with the environment.
The system built for this project consisted of a two-dimensional laser scanner mounted on a rotating shaft. Shaft was rotated giving the scanner a nodding motion and therefore making possible to scan a three-dimensional point cloud. The point cloud was used for obstacle position estimation and tracking using tools provided by Point Cloud Library. The system performance was evaluated using a physical object whose position was estimated using the scanner, and moving the object in a controlled and measurable manner.
The system tested within this project was able to track obstacle location. The error in obstacle position was up to 0.15 m, and tracking was delayed up to 0.5 s. The position estimation also tended to have high sudden variations not related to the real movement of the obstacle. The performance was not quite what modern hardware used in similar tasks is capable of, and suggests that either the approach presented here is not optimal, or that there are several areas that have to be improved. The issue with high variation in the position estimate must also be investigated should this line of research be continued in the future.
The system built for this project consisted of a two-dimensional laser scanner mounted on a rotating shaft. Shaft was rotated giving the scanner a nodding motion and therefore making possible to scan a three-dimensional point cloud. The point cloud was used for obstacle position estimation and tracking using tools provided by Point Cloud Library. The system performance was evaluated using a physical object whose position was estimated using the scanner, and moving the object in a controlled and measurable manner.
The system tested within this project was able to track obstacle location. The error in obstacle position was up to 0.15 m, and tracking was delayed up to 0.5 s. The position estimation also tended to have high sudden variations not related to the real movement of the obstacle. The performance was not quite what modern hardware used in similar tasks is capable of, and suggests that either the approach presented here is not optimal, or that there are several areas that have to be improved. The issue with high variation in the position estimate must also be investigated should this line of research be continued in the future.