Keywords
circuit breaker test;robot wiring;vision location;background data augmentation;improved YOLOv4
Abstract
In order to improve the accuracy and reliability of circuit breaker test robot wiring, a locating method with background augmentation and improved YOLOv4 on the basis of binocular vision and deep learning object detection technology is proposed in this paper. Background mixed shear method is adopted in the proposed method to solve the low generalization ability and accuracy problems caused by insufficient training background features. Therefore, the accuracy and reliability of wiring under the background disturbance such as different test sites and people walking are increased. Furthermore, the backbone of YOLOv4 is replaced to Mobiledets to optimize the reasoning period of the object detection model. So that the efficiency of robot wiring is improved. Simulation and test results show that the accuracy of detection model based on the proposed method is 99.9%, the robot wiring accuracy is 98.8%, and the wiring time is reduced by 57 s. Comparison and analysis indicate that, the method proposed in this paper is superior to other methods in robot wiring accuracy and time, which can provide technical support for the practicability of breaker robot test platform.
DOI
10.19781/j.issn.1673-9140.2023.02.022
First Page
196
Last Page
204,239
Recommended Citation
HE, Shenghong; WU, Xiaoping; WANG, Junbo; and ZHANG, Yin
(2023)
"Wiring locating method for circuit breaker test robot based on background augmentation and improved YOLOv4,"
Journal of Electric Power Science and Technology: Vol. 38:
Iss.
2, Article 22.
DOI: 10.19781/j.issn.1673-9140.2023.02.022
Available at:
https://jepst.researchcommons.org/journal/vol38/iss2/22