Machine Vision-Based Product Classification on the Prototype Production Line

V sklopu diplomske naloge smo naredili projekt, pri katerem smo s pomočjo strojnega vida razvrščali izdelke na liniji. Industrijska kamera je na podlagi zaznavanja obrisov izdelkov komunicirala s testno enoto, ta pa je delovala na podoben način kot prava proizvodna linija. S pomočjo programirljivega logičnega krmilnika smo se z dobljenimi informacijami odločili, katere operacije naj se izvršijo za razvrščanje izdelkov in njihovih palet. Poleg zajemanja slik in programiranja procesnega krmilnika smo preučili vsa orodja, ki smo jih uporabili in tudi vso potrebno strojno opremo. Glede na to, da je veliko načinov, s katerimi se proizvodnje ukvarjajo s pridobivanjem informacij o izdelkih na linijah, smo preko projekta ugotovili, da je najbolj praktičen primer detekcije izdelkov s pomočjo kamere in strojnega vida. Palete, ki nosijo izdelke, so izpostavljene velikim temperaturam in neskončnim ciklom kroženja po liniji. To pa lahko prinese veliko problemov za detekcijo pri drugačnih metodah. S pomočjo strojnega vida, smo pri projektu dobili zelo dobre rezultate zaznavanja obrisov, s tem pa pravilno delovanje projekta.We made a project in which we sorted products on the industrial production line using machine vision. The industrial camera communicated with the test unit based on the detection of the outlines of the product which acted in a similar way to a real production line. With the help of a programmable logic controller and the information obtained from teh camera we decided which operations to perform to sort the products and their pallets. In addition to capturing images and programming the process controller, we looked at all the tools we used and all the harware we needed. Considering that there are many ways in which manufacturers obtain product information on the production lines, we have found through the project that the most practical example of product detection is through camera and machine vision. Product-carrying pallets are exposed to high temperatures and endless running cycles. This, however, can lead to many detection problems with different methods. With the help of machine vision, the project got very good outline detection results, which made the project work properly