A Meta-Learning-Driven Joint Detection and Localization Algorithm for Low-Voltage Electric Meters

With the development of smart grids, the demand for intelligent detection has surged in large-scale power monitoring and management systems. Low-voltage electricity meters, widely used in low-voltage power systems such as residential communities and office buildings, are typically installed in conce...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Proceedings (International Conference on Computer Engineering and Applications. Online) s. 01 - 05
Hlavní autori: Huang, Zhongmou, Li, Hualin, Lu, Jiyang, Zhao, Tingkang, Long, Guigan, Huang, Aili, Chen, Zixin
Médium: Konferenčný príspevok..
Jazyk:English
Vydavateľské údaje: IEEE 25.04.2025
Predmet:
Accuracy
Adaptation models
Detection
Electricity
Feature extraction
Lighting
Location awareness
Low voltage
Low-Voltage Electric Meters
Meta-Learning
Metalearning
Meters
Robustness
YOLO
ISSN:2159-1288
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:With the development of smart grids, the demand for intelligent detection has surged in large-scale power monitoring and management systems. Low-voltage electricity meters, widely used in low-voltage power systems such as residential communities and office buildings, are typically installed in concealed locations like electrical cabinets in corridors or elevator shafts. Their detection and localization are often hindered by obstructions, insufficient lighting, and viewing angles. Existing detection algorithms require extensive annotated data for training and suffer from reduced accuracy and efficiency due to on-site environmental variations, such as camera angles and lighting conditions. To address these challenges, this paper proposes Meta-Real Time-Detect (Meta-RD), a meta-learningintegrated object detection framework. By incorporating feature weighting to enhance detection accuracy, localization precision, and algorithmic robustness, Meta-RD achieves a highly efficient model with strong adaptability using minimal training data. Comparative experiments demonstrate that the proposed MetaRD architecture improves accuracy by 3 % compared to prior methods and exhibits superior resilience to environmental changes.
ISSN:2159-1288
DOI:10.1109/ICCEA65460.2025.11103386