Journal
MEASUREMENT
Volume 178, Issue -, Pages -Publisher
ELSEVIER SCI LTD
DOI: 10.1016/j.measurement.2021.109324
Keywords
Surface defects; Target detection algorithm; Convolutional neural network; Deep learning
Funding
- Dongguan social science and technology development (key) project [DongkeTong [2020]77, 2020507156157]
Ask authors/readers for more resources
This paper proposes a method for detecting and identifying commutator surface defects, by applying YOLOv3 target detection to commutator surface defect detection and recognition, designing a network with smaller model size, fewer parameters, and faster running time, and achieving good accuracy in experiments.
With low accuracy and poor efficiency, the manual and traditional detection of commutator surface defects cannot meet efficiency and timeliness requirements. Thus, a method is proposed to detect and identify commutator surface defects. YOLOv3 target detection is applied to commutator surface defect detection and recognition, and the model size and parameter number are reduced without significantly reducing detection accuracy. First, this paper proposes a separable residual module based on deep separable convolutions and residual networks. A network with shallower layers and fewer channels is designed for quick detection and recognition of commutator surface defects. Finally, the algorithm herein is evaluated on SD_data. The experimental results indicate that while maintaining good accuracy, this method has a smaller model size, fewer parameters and faster running time than the YOLOv3 network. Moreover, accuracy is improved to a certain extent compared to the current commutator surface defect detection methods, so it can be better applied to real-time detection and recognition of commutator surface defects.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available