Infrared Image Recognition of Power Equipment Using Improved YOLOv4

Table 1. Parameter value of MSRCR algorithm
View table
Table 1. Parameter value of MSRCR algorithm
Parameter N C₁ C₂ C₃ θ₁ θ₂ θ₃ $κ$ $σ$
Value 3 15 75 225 0.33 0.33 0.34 43 131

Table 2. Number of devices on infrared image dataset
View table
Table 2. Number of devices on infrared image dataset
Parameter Insulator Bushing Disconnector Breaker Arrester Voltage transformer Current transformer Transformer
Number 496 451 418 401 405 382 334 313

Table 3. Test results of the proposed method on different power equipments
View table
Table 3. Test results of the proposed method on different power equipments
Electric power equipment AP /% Speed /（frame·s^-1）
Insulator 97.15 71
Bushing 96.57
Disconnector 97.83
Breaker 98.23
Arrester 96.51
Voltage transformer 94.62
Current transformer 96.35
Transformer 93.24
mAP /% 96.31

Table 4. Performance comparison of different methods

View table

Table 4. Performance comparison of different methods

Method	Raw infrared image		Image enhanced by MSRCR algorithm
Method	mAP@0.5 /%	Speed /（frame·s^-1）	mAP@0.5 /%	Speed /（frame·s^-1）
Faster R-CNN	93.13	16	94.03	17
SSD	90.13	33	91.37	35
YOLOv3	91.54	65	92.71	68
YOLOv4	92.73	74	93.69	75
Proposed method	95.12	69	96.31	71

Table 5. Comparison of experimental accuracy under different parameters of Focal loss
View table
Table 5. Comparison of experimental accuracy under different parameters of Focal loss
Modulation parameter mAP /%
β=0.5，γ=1 92.63
β=0.7，γ=1 92.66
β=0.9，γ=1 92.82
β=0.9，γ=2 94.73
β=0.9，γ=3 96.31
β=0.9，γ=4 95.26
β=0.9，γ=5 94.67
β=0.7，γ=3 96.18

Tools

Get Citation

Copy Citation Text

Zhongxing Duan, Yuming Zhang, Jiahao Ma. Infrared Image Recognition of Power Equipment Using Improved YOLOv4[J]. Laser & Optoelectronics Progress, 2022, 59(24): 2410002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites