We present a novel smart camera - the FlexSight C1 - designed to enable an industrial robot to detect and localize several types of objects and parts in an accurate and reliable way. The C1 integrates all the sensors and a powerful mini computer with a complete Operating System running robust 3D reconstruction and object localization algorithms on-board, so it can be directly connected to the robot that is guided directly by the device during the production cycle without any external computers in the loop. In this paper, we describe the FlexSight C1 hardware configuration along with the algorithms designed to face the model based localization problem of textureless objects, namely: (1) an improved version of the PatchMatch Stereo matching algorithm for depth estimation; (2) an object detection pipeline based on deep transfer learning with synthetic data. All the presented algorithms have been tested on publicly available datasets, showing effective results and improved runtime performance.
FlexSight - A Flexible and Accurate System for Object Detection and Localization for Industrial Robots / Evangelista, Daniele; Imperoli, Marco; Menegatti, Emanuele; Pretto, Alberto. - (2019), pp. 58-63. (Intervento presentato al convegno IEEE International Workshop on Metrology for Industry 4.0 and IoT tenutosi a Napoli; Italy) [10.1109/METROI4.2019.8792902].
FlexSight - A Flexible and Accurate System for Object Detection and Localization for Industrial Robots
Evangelista, DanielePrimo
;Imperoli, Marco
;Pretto, Alberto
Ultimo
2019
Abstract
We present a novel smart camera - the FlexSight C1 - designed to enable an industrial robot to detect and localize several types of objects and parts in an accurate and reliable way. The C1 integrates all the sensors and a powerful mini computer with a complete Operating System running robust 3D reconstruction and object localization algorithms on-board, so it can be directly connected to the robot that is guided directly by the device during the production cycle without any external computers in the loop. In this paper, we describe the FlexSight C1 hardware configuration along with the algorithms designed to face the model based localization problem of textureless objects, namely: (1) an improved version of the PatchMatch Stereo matching algorithm for depth estimation; (2) an object detection pipeline based on deep transfer learning with synthetic data. All the presented algorithms have been tested on publicly available datasets, showing effective results and improved runtime performance.File | Dimensione | Formato | |
---|---|---|---|
Evangelista_Postprint_FlexSight_2019.pdf
Open Access dal 13/08/2020
Note: https://ieeexplore.ieee.org/abstract/document/8792902
Tipologia:
Documento in Post-print (versione successiva alla peer review e accettata per la pubblicazione)
Licenza:
Tutti i diritti riservati (All rights reserved)
Dimensione
6.05 MB
Formato
Adobe PDF
|
6.05 MB | Adobe PDF | |
Evangelista_FlexSight_2019.pdf
solo gestori archivio
Tipologia:
Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza:
Tutti i diritti riservati (All rights reserved)
Dimensione
252.1 kB
Formato
Adobe PDF
|
252.1 kB | Adobe PDF | Contatta l'autore |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.