This work addresses an automated system that supports the "Maintenance on Condition"paradigm. The system automatically processes auto-recorded video streams collected by a wayside camera capture device, and proceeds to a preliminary analysis in search for posture errors.We present and discuss a novel system that provides a proper identification of pantograph heads from video sources or image sequences and proceeds to recover the 3D asset from a single 2D image. The system exploits modern Deep Learning (DL) networks to extract a preliminary pantograph shape from the video sequence, then applies an optimization algorithm to estimate accurate position and orientation.A comparative and exhaustive analysis with few hundreds of samples shows the accuracy of the algorithm and the robustness to changes in illumination and other environmental conditions.

Accurate Identification of 3D Pose through Reprojection onto a Single Image from Mask-RCNN Contour

Baris G.
;
Avizzano C. A.
2020-01-01

Abstract

This work addresses an automated system that supports the "Maintenance on Condition"paradigm. The system automatically processes auto-recorded video streams collected by a wayside camera capture device, and proceeds to a preliminary analysis in search for posture errors.We present and discuss a novel system that provides a proper identification of pantograph heads from video sources or image sequences and proceeds to recover the 3D asset from a single 2D image. The system exploits modern Deep Learning (DL) networks to extract a preliminary pantograph shape from the video sequence, then applies an optimization algorithm to estimate accurate position and orientation.A comparative and exhaustive analysis with few hundreds of samples shows the accuracy of the algorithm and the robustness to changes in illumination and other environmental conditions.
2020
978-1-7281-8956-7
File in questo prodotto:
File Dimensione Formato  
ETFA20_baris_avizzano.pdf

solo utenti autorizzati

Tipologia: Documento in Post-print/Accepted manuscript
Licenza: Non pubblico
Dimensione 3.89 MB
Formato Adobe PDF
3.89 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

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/11382/535166
 Attenzione

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

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
social impact