Vision-based Explainable Action Recognition Model for Polluted Soil Tracking - Université de technologie de Troyes Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Vision-based Explainable Action Recognition Model for Polluted Soil Tracking

Résumé

On an excavation site, it is difficult to keep track of all the movements of excavated soil, often resulting in the loss or mixing of soil. This leads to additional costs and difficulties in re-using this soil. The diversity of work sites and the lack of effective markers make it difficult for an operator to monitor the soil. The aim of this research is to propose a versatile, operator-independent solution to this problem, based on camera networks and action recognition. Since it is difficult to track soil directly, this study focuses on the machines that extract and move it. This research proposes the construction of a convolutional neural network associated with computer vision methods to detect earthmoving machinery and recognize the actions in progress. By considering the operating protocols of a construction site, this method aims to allow the tracking of soil by interposed vehicles in real time. The results obtained by this model allow to efficiently detect operating machines as well as their main actions.
Fichier non déposé

Dates et versions

hal-04447100 , version 1 (08-02-2024)

Identifiants

Citer

Edern Viannay, Benitha Vasseur, Hichem Snoussi, Dominique Poux. Vision-based Explainable Action Recognition Model for Polluted Soil Tracking. 2023 International Conference on Control, Automation and Diagnosis (ICCAD), May 2023, Rome, Italy. pp.1-5, ⟨10.1109/ICCAD57653.2023.10152357⟩. ⟨hal-04447100⟩
8 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More