Tyre brand, its size, model, age and condition monitoring are critical for many vehicle users. The detection and the recognition of plastic components defects result essential. Image classification has become one of the key applications in image processing and computer vision domain. It has been used in several fields such as medical area and intelligent transportation. Recently, results of deep neural networks (DNN) foreshadow the advent of reliable classifiers to perform such visual tasks. DNNs require learning of many parameters from raw images; hence, several images with class annotations are needed. These images are very expensive since pixel-level annotations are required. In this paper, we introduce a deep learning approach to detect and classify five classes of plastic components defects. A novel dataset of tyre images is collected and the images are manually labelled. The experiments are conducted on this dataset by comparing the performances of three DNNs such as UNet, FPN and LinkNet. Results yield high values of F1-score and show the effectiveness and the suitability of the proposed approach.

Detection and Classification of Defects in Plastic Components Using a Deep Learning Approach / Mameli, M.; Paolanti, M.; Mancini, A.; Frontoni, E.; Zingaretti, P.. - 412 LNNS:(2022), pp. 713-722. [10.1007/978-3-030-95892-3_53]

Detection and Classification of Defects in Plastic Components Using a Deep Learning Approach

Mameli M.;Paolanti M.;Mancini A.;Frontoni E.;Zingaretti P.
2022-01-01

Abstract

Tyre brand, its size, model, age and condition monitoring are critical for many vehicle users. The detection and the recognition of plastic components defects result essential. Image classification has become one of the key applications in image processing and computer vision domain. It has been used in several fields such as medical area and intelligent transportation. Recently, results of deep neural networks (DNN) foreshadow the advent of reliable classifiers to perform such visual tasks. DNNs require learning of many parameters from raw images; hence, several images with class annotations are needed. These images are very expensive since pixel-level annotations are required. In this paper, we introduce a deep learning approach to detect and classify five classes of plastic components defects. A novel dataset of tyre images is collected and the images are manually labelled. The experiments are conducted on this dataset by comparing the performances of three DNNs such as UNet, FPN and LinkNet. Results yield high values of F1-score and show the effectiveness and the suitability of the proposed approach.
2022
Lecture Notes in Networks and Systems
978-3-030-95891-6
978-3-030-95892-3
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11566/315912
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact