20.500.12556/DKUM-64503 Sledenje gibanju žogice za namizni tenis s postopki računalniškega vida na osnovi dveh kamer in odziva robota Tracking the movement of table tennis ball by using computer vision methods based on two cameras and response of robot V diplomskem delu smo se ukvarjali z razvojem robotsko-računalniškega sistema, s katerim smo predvsem preučevali postopke računalniškega vida z namenom, da robotu povečamo umetno inteligenco. Načrtali smo preprostega robota za omenjeno igranje namiznega tenisa in ga nadgradili z lastno izdelanim sistemom računalniškega vida. S pomočjo dveh kamer in generiranja stereo slike smo s postopki računalniškega vida zaznali žogico za namizni tenis, ji sledili, z rekonstrukcijo 3D geometrijskega položaja pa smo uvedli še model predikcije njene končne poti. Na osnovi predvidene poti žogice nato premaknemo robota do položaja, v katerem žogico lahko zadane. V tem diplomskem delu smo uporabili preproste metode računalniškega vida, obdelave slik in krmiljenja robota, ki niso dovolj zmogljivi, natančni in odzivni za igranje t. i. »resne« igre namiznega tenisa proti robotu. Naša rešitev je primerna zgolj za podajanje žogice z robotom. S pomočjo večih eksperimentov smo potrdili pravilnost modela za predvidevanje poti žogice, izmerili smo natančnost 3D rekonstrukcije položaja žogice, modelirali pa smo tudi funkcijo, s pomočjo katere ocenimo čas, potreben za premik robota na izbran položaj. Na koncu smo ocenili in izmerili robne pogoje, v katerih se naš robot še ustrezno odzove na žogico. In this thesis, we worked towards the development of a computer-robot system where we focused our study on computer vision processes in order to increase a robot's artificial intelligence. We programed a simple robot to play table tennis and then enhanced it with its own developed computer vision system. With the help of two cameras and the generation of stereo images, we used computer vision procedures to detect a table tennis ball, track the table tennis ball, and introduced, using a 3D reconstruction of the tracked ball's geometric position, a prediction model to predict the final position of the table tennis ball. The robot moves to the position, based on the ball's predictive path, where it needs to be so that it hits the ball. We used simple computer vision, image processing and robot control methods that, while not powerful, accurate or responsive enough to play a "serious" game of table tennis against the robot, enabled us to pass balls to the robot. During the course of several experiments, we confirmed the efficacy of the ball trajectory prediction model, measured the precision of the position of the 3D reconstructed table tennis ball, and modelled a function that estimated the time required to move the robot to the selected position. Finally, we evaluated and measured the boundary conditions under which our robot could respond to the table tennis ball. računalniški vid robotika robotsko-računalniški sistem namizni tenis predikcija computer vision robotics computer-robot system table tennis prediction true true false G. Mlakar Slovenski jezik Angleški jezik Diplomsko delo/naloga 2016-10-11 13:31:02 2016-12-02 14:26:11 2022-07-05 03:08:01 2019-07-11 14:03:59 2016 0 [Maribor 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.9(043.2) 20315414 URN:SI:UM:DK:41ZOWWOA VS_Mlakar_Gregor_2016.pdf VS_Mlakar_Gregor_2016.pdf 1 ECB3EC6FAC072255209904BADB42398C e000a4ac5cd07b27afb55524fe8077104a65dde4bff81cced922f42a2c493a70 e33bb1c5-7c0d-11eb-bb7a-00155d0001ca https://dk.um.si/Dokument.php?lang=slv&id=108621 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0