Detekcija gibanja robota s 3D kamero

diplomsko delo

Domen Vovk (Author), Marko Munih (Mentor), Luka Pogačnik (Co-mentor)

Abstract

Diplomsko delo obravnava sledenje robotu z uporabo oblaka točk in določanje kotov v sklepih robota. Sestavili smo sistem z LIDAR-jem in robotom ter napisali program, ki je sposoben slediti premikom robota. Z globinsko kamero smo med izvajanjem robotskega programa zajemali oblake točk in nanje z algoritmom iterativne najbližje točke prilegali modela dveh glavnih robotskih segmentov. Kote v oseh robota smo določili na podlagi transformacijskih matrik, ki jih je algoritem vrnil. Testirali smo vplive različnih hitrosti izvajanja robotskega programa in preverili dva različna položaja kamere glede na robota. Ugotovili smo, da poravnava deluje tako pri počasnih kot hitrejših premikih robota. Program daje slabe rezultate le v primeru, ko LIDAR ne vidi ustreznih segmentov robota.

Keywords

globinski vid;iterativna najbližja točka;oblak točk;detekcija robota;univerzitetni študij;Elektrotehnika;diplomske naloge;

Data

Language:	Slovenian
Year of publishing:	2022
Typology:	2.11 - Undergraduate Thesis
Organization:	UL FE - Faculty of Electrical Engineering
Publisher:	[D. Vovk]
UDC:	004.93:007.52(043.2)
COBISS:	120455683
Views:	22
Downloads:	5
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Robot motion detection with a 3D camera
Secondary abstract:	The thesis deals with robot tracking using a point cloud and calculation of the robot's joint angles. We assembled a system with a LIDAR and a robot and created a program capable of robot tracking. While the robot was moving, a point cloud was repeatedly captured by the depth camera. Using the iterative closest point algorithm, we could then align the models of the two main segments of the robot with the captured point cloud. The joint angles were calculated using the transformation matrix returned by the algorithm. We tested our program with different speeds of the robot and two different camera positions. The experimental results show that the software functions properly with slow robot movements and gives good resuts with higher speeds as well. Errors occur when the segments cannot be seen by the LIDAR.
Secondary keywords:	Depth vision;iterative closest point;point cloud;robot detection;
Type (COBISS):	Bachelor thesis/paper
Study programme:	1000313
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Fak. za elektrotehniko
Pages:	VIII, 58 str.
ID:	16411038