Keywords
object detection, deep learning, YOLOv3, scene recognition, intersection over union
Abstract
Power grid are often operated in dangerous scenes such as high altitudes and high voltages. The scenes pose threats to the safety of electric power operators. Relying only on human supervision alone often leads to inadequate supervision. Existing target detection algorithms can only perform simple safety identification and can not identify illegal operations based on specific power operation scenarios. To solve this problem, this paper proposes an illegal operation recognition algorithm based on YOLOv3 in the specific power operation scenario. The YOLOv3 algorithm is selected for target detection incorporating the scene recognition mechanism contemporarily. The logic judgment function is set by reference to the intersection over union to detect the violation of power operations in specific scenarios. After taking the welding scene as an example for experimental verification, the results show the detection accuracy of this model is calculated to be 82.15%, which proves the effectiveness of the method. Meanwhile, this paper also puts forward several suggestions for subsequent optimization of the model.
DOI
10.19781/j.issn.1673-9140.2021.03.024
First Page
195
Last Page
202
Recommended Citation
Qiu, Hao; Zhang, Wei; Peng, Boya; Ding, Zhaojun; and Lin, Xiangyu
(2021)
"Illegaloperation recognition algorithm based on YOLOv3 in specific power operation scenario,"
Journal of Electric Power Science and Technology: Vol. 36:
Iss.
3, Article 24.
DOI: 10.19781/j.issn.1673-9140.2021.03.024
Available at:
https://jepst.researchcommons.org/journal/vol36/iss3/24