Razvoj sistema za kontrolo zvarov na podlagi strojnega vida

ŽUFIČ, PRIMOŽ (author), Štruc, Vitomir (mentor)

V diplomskem delu je predstavljen razvoj in izdelava sistema strojnega vida za kontrolo zvarov kratkostičnih obročkov na statorjih motorjev z zasenčenimi poli v podjetju Ydria Motors d.o.o. V prvem delu so predstavljene zahteve, ki jih mora sistem izpolnjevati. Sledi teoretična podlaga o kamerah in objektivih, ki se uporabljajo za namene strojnega vida. Diplomsko delo se nadaljuje s poglavjem o osnovnih tipih osvetlitev, ki se uporabljajo za namene strojnega vida. V glavnem delu je predstavljen celoten projekt od zasnove pa do končnega izdelka in testiranj. Sistem je sestavljen iz dveh kamer DMK 42AUC03 proizvajalca The Imaging Source. Ena izmed kamer zajema sliko statorskega paketa z vrha in uporablja obročasto LED (light emitting diode) osvetlitev, druga kamera pa zajema sliko od strani, pri čemer uporablja LED osvetlitev, usmerjeno proti kameri. Prva osvetlitev je bila izdelana in razvita namensko, pri drugi pa gre za običajno komercialno izvedbo. Uporabljen je računalnik Raspberry Pi 2, na katerem deluje strogo namenski program, napisan v programskem jeziku Python. Kot glavna podpora za obdelavo zajetih slik je uporabljena knjižnica OpenCV. Raspberry je priključen na namensko vezje, ki zagotavlja pretvorbo vhodnih in izhodnih nivojev na 24 V DC. Vezje skrbi za vklop osvetlitev in komunikacijo s programirljivim logičnim krmilnikom. Vsi glavni elektronski deli so nameščeni v kovinski omarici, na ogrodju stroja, sistem za zajem slike pa na primernem mestu v kovinskem ohišju znotraj obdelovalnega dela stroja. S pomočjo testiranj na testni bazi in na stroju samem je bilo ugotovljeno, da sistem zadovoljivo služi svojemu namenu. Možnosti za izboljšave so odprte predvsem pri obdelavi zajetih slik. Fizični in programski del po nekaj odpravljenih težavah delujeta zanesljivo v času celotne izmene.

• Language: Slovenian
• Keywords: zvar, kamera, objektiv, Raspberry Pi, Python, OpenCV, osvetlitev, klasifikacija
• Work type: Bachelor thesis/paper
• Organization: FE - Faculty of Electrical Engineering
• Year: 2016
• PID: 20.500.12556/RUL-81162
• Publication date in RUL: 30.03.2016

Secondary language

• Language: English
• Title: Development of a machine vision system for weld quality control

Abstract

In this thesis we present the development and subsequent production of a machine vision system for quality control of welds. The welds are applied on a specific part of a shadow pole motor produced by the Ydria Motors company. In the first section we present the system requirements. Next we describe the camera and lens used in machine vision applications and elaborate on exisitng sources of illumination in machine vision systems. In the main part of the thesis the entire project is presented and the developed system is described. The system consists of two cameras (The Imaging Source DMK 42AUC03). One of the cameras is used to acquire images of the top view and the other images of the side view of the weld. We use LED light arranged into a ring structure for the top view and LED backlight for the side view. The top view lighting was designed specially for the system, while the second one is available commercially. We use a Raspberry Pi 2 running custom software (written in Python) for the computational part of the system. For image processing functionality we rely on the OpenCV library. The Raspberry PI 2 is connected to a specially designed printed circuit board which ensures the conversion of input and output signals to 24 V DC. The curcuitry is used to switch the lighting on and off and to communicate with a programmable logic controller. All of the main electronic parts are mounted in a metal box attached to the frame of the system. The image acquisition system is placed in a appropriate position inside a metal casing. Based on quantitative results obtained on a database of test images acquired in working conditions, it was determined that the system performs satisfactory. There is still room for improvement mainly in the processing of captured images. After a few successfully solved problems, the physical and software parts are working reliably throughout the wroking day.

• Keywords: weld, camera, lens, Raspberry Pi, Python, OpenCV, lighting, classification