Usefulness of Shape-From-Shading Algorithm for Surface Defect Detection

V tem delu nas je zanimalo trenutno stanje področja rekonstrukcije oblike iz senčenja in njegov potencial za uporabo v industrijskih okoljih za zaznavanje nepravilnosti na površini. Za preizkušanje smo izbrali odprtokodni algoritem od priznanih avtorjev iz področja. Naredili smo testno okolje, ki zadovolji predpostavkam algoritma ter slikali več različnih predmetov. Med preizkušanjem smo naleteli na težave z izbiro pravilnih vrednosti parametrov. Problematična se je izkazala tudi poravnava predhodnega modela s sliko. Te probleme smo naslovili in po ustreznih prilagoditvah se je algoritem izkazal za presenetljivo dobrega. Vsi rezultati so bili uporabni z vidika razpoznave nepravilnosti in nakazujejo dober potencial za industrijsko uporabo.This work explores the potential of current shape from shading methods for error detection in industrial environments. A state of the art algorithm from reputable authors was chosen for testing. An environment that satisfies the authors requirements was constructed and several different items were sampled. The issue of choosing the correct set of parameters, as well as aligning the prior with the captured picture, was encountered. After addressing the issue and making some minor adjustments, the results turned out surprisingly well. All of them are useful from an error recognition standpoint and hold promise for industrial use