Contour tracking by robot

Cilj diplomskog rada je istražiti mogućnosti detekcije kontura lima i praćenje pomoću robotske ruke. Diplomski rad je podijeljen prema redoslijedu zadataka koji su se rješavali kako bi robot što točnije mogao pratiti konturu lima. Za ostvarivanje praćenja konture bilo je potrebno osigurati radni prostor, robota, kameru i računalo. Prvi cilj zadatka je primijeniti 2D vizijski sustav za određivanje kontura predmetanepoznatog oblika. U tu svrhu razvijen je računalni program za analizu radnog prostora putem vizijskog sustava u C++ programskom jeziku. Kao podloga programu korišteni su algoritmi OpenCV baze algoritama za računalni vid i umjetnu inteligenciju. Zatim je bilo potrebno kalibrirati kameru radi dobivanja odnosa piksela na slici i milimetara u realnom 3D prostoru, te odstranjivanja distorzija na fotografijama. Također, pomoću matrica homogenih transformacija povezani su koordinatni sustavi robota i kamere. Kako bi robot mogao kontinuirano pratiti točke konture na njemu se nezavisno odvijaju Teach Pendant program koji je zadužen za gibanje robota i Karel program za primanje podataka o točkama konture sa računala. Sustav je rezultirao vrlo preciznim praćenjem konture lima te su na kraju rada dani savjeti za moguću implementaciju u realno industrijsko okruženje.The aim of this master's thesis is to explore the possibilities of sheet metal contour detection and tracking with robotic arm. Thesis is divided according to the sequence of tasks needed to be solved in order for robot to track the contour accurately. For contour tracking goal it is needed to secure the working area, robotic arm, industrial camera and computer. First task of the thesis was to apply a 2D vision system for contour detection. For that purpose computer program for work space analysis in C++ programming language was developed. Program was built around OpenCV library alghorithms for computer vision and artificial intelligence. Than it was needed to calibrate the camera in order to get the relation between pixels on the image and milimeters in 3D space. Another goal of the calibration was to removedistortions on the images from camera. Additionaly, coordinate systems of camera and robot were connected with transformation matrices. In order for robot to be able to track the contour continuously, Teach Pendant program responsible for robot movement and Karel program for receiving of data about sheet metal contours from computer ran independently one to another. The system resulted with accurate sheet metal contour tracking. At the end of the thesis were given the advices for possible implementation into the real industrial surrounding