Operator Action Recognition Based on Improved ST-GCN

With the rapid development of the electric power industry, the safety problems of the personnel working with electricity are becoming more and more prominent. Traditional solutions relying on manual methods perform poorly in ensuring safety and are in urgent need of updating and improvement. In rece...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:International Conference on Communications, Information System and Computer Engineering (Online) S. 419 - 422
Hauptverfasser: Fan, Longxin, Sun, Hongbin
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 10.05.2024
Schlagworte:
CBAM
Electricity
Graph convolutional networks
Heuristic algorithms
human skeleton
Manuals
motion recognition
OpenPose
Pose estimation
Power systems
Safety
ST-GCN
ISSN:2833-2423
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:With the rapid development of the electric power industry, the safety problems of the personnel working with electricity are becoming more and more prominent. Traditional solutions relying on manual methods perform poorly in ensuring safety and are in urgent need of updating and improvement. In recent years, the application of computer vision technology has provided new possibilities for improving safety standards, especially the important role of human skeletal dynamics in action recognition is becoming more and more prominent. In order to effectively solve the existing safety challenges of power-carrying operations, this paper proposes a power-carrying operation action recognition method based on ST-GCN (Spatio-Temporal Graph Convolutional Networks). The method combines a pose estimation algorithm and an action recognition algorithm to achieve the purpose of action recognition, in which the improved OpenPose algorithm is used for pose estimation to obtain the skeletal data, and then the ST-GCN is used for action recognition, in which a channel attention module CBAM (Convolutional Block Attention Module) is added for enhancement of action features. After experiments, it is confirmed that the spatio-temporal graph convolutional network method proposed in this paper shows excellent results in practical applications and effectively improves the safety and security level of the personnel working with electricity.
ISSN:2833-2423
DOI:10.1109/CISCE62493.2024.10653348