Detekcija prask na pokrovih avtomobilskih koles z računalniškim vidom

Zupanič, Matjaž (author), Kristan, Matej (mentor)

Detekcija prask na platiščih avtomobilov s klasičnimi pristopi računalniškega vida ne daje dobrih rezultatov. Zato, ker so platišča različnih oblik, narejena iz različnih materialov in raznih barv. Tudi praske so različnih oblik, barv in velikosti, večrkrat tudi slabo vidne. Na platiščih se pojavlja tudi umazanija, ki dodatno ovira zaznavo. Zaradi tega je potrebna uporaba močnejših orodij. To so konvolucijska nevronska omrežja, ki doživljajo bliskovit razvoj nekaj zadnjih let. V diplomski nalogi analiziramo možnost zaznavanja prask na platiščih z globoko nevronsko mrežo. Za potrebe zaznave je potrebna segmentacija vsake točke vhodne slike. Ker so praske v primerjavi s celotnim platiščem majhne, smo se odločili uporabiti polno konvolucijsko omrežje U-Net. V namen diplome je bila pripravljena zbirka označenih fotografij, ki je lahko izhodišče za nadaljnje raziskave. Razvit model uspešno detektira praske, kljub mali učni množici. Na nevideni testni množici pripravljeni le za namen evalvacije je po metodi mIoU dosegel točnost 62,8\,\%. V primeru izboljšav in dodelav pa bi naš detektor prask bil primeren tudi za industrijsko rabo.

• Language: Slovenian
• Keywords: konvolucijska nevronska omrežja, U-Net, segmentacija, platišča
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2021
• PID: 20.500.12556/RUL-130327
• COBISS.SI-ID: 77571331
• Publication date in RUL: 13.09.2021

ZUPANIČ, Matjaž, 2021, Detekcija prask na pokrovih avtomobilskih koles z računalniškim vidom [online]. Bachelor’s thesis. [Accessed 24 March 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=130327

Secondary language

• Language: English
• Title: Scratch detection on car wheel covers using computer vision

Abstract

Detection of scratches on car rims with classical computer vision approaches does not produce good results, because the rims are of different shapes, made of different materials and in different colors. Scratches are also in different shapes, colors and sizes. Scratches are often poorly visible and there is also dirt on rims, which further impedes visibility of them. This requires the use of more powerful tools, like convolutional neural networks that have been experiencing rapid development over the last few years. In thesis we analyze possibility of scratch detection on car rims with deep neural network. Segmentation of each point in the input image is required for detection purposes. Because the scratches are small compared to the entire region of the rim, we decided to use a fully convolutional network U-Net. For a purpose of the thesis, a collection of annotated pictures was prepared, which can be a starting point for further research. The developed model successfully detects scratches, despite the small learning set. On an unseen test set prepared for evaluation purposes only, it achieved 62.8\,\% accuracy using the mIoU method. With further improvements and refinements, our scratch detector would also be suitable for industrial use.

• Keywords: convolutional neural networks, U-Net, segmentation, rims