Insulator Defect Detection Based on Multi-Scale Feature Coding and Dual Attention Fusion

Lirong Li; Peng Chen; Yunliang Zhang; Kai Zhang; Wei Xiong; Pengcheng Gong

doi:10.3788/LOP202259.2410008

Laser & Optoelectronics Progress, Volume. 59, Issue 24, 2410008(2022)

Insulator Defect Detection Based on Multi-Scale Feature Coding and Dual Attention Fusion

Lirong Li^1,2、*, Peng Chen¹, Yunliang Zhang¹, Kai Zhang¹, Wei Xiong^1,2, and Pengcheng Gong^1,2

Author Affiliations

¹School of Electrical and Electronic Engineering, Hubei University of Technology, Hubei 430064, Wuhan, China

²Hubei Engineering Research Center of New Energy and Power Grid Equipment Safety Monitoring, Hubei 430064, Wuhan, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(15)

Fig. 1. CenterNet structure

Download full size

Fig. 2. Structure of proposed algorithm

Download full size

Fig. 3. Schematic diagrams of Bottleneck block and Res2Net Module. (a) Bottleneck block; (b) Res2Net module

Download full size

Fig. 4. ASPP module

Download full size

Fig. 5. Schematic diagram of dual attention fusion

Download full size

Fig. 6. Loss curve

Download full size

Fig. 7. Sample dataset

Download full size

Fig. 8. Example of real box labels

Download full size

Fig. 9. Visualization of detection results. (a) Visualization of insulator detection results; (b) visualization of test results of insulators and defective insulators

Download full size

Table 1. Definitions of TP, TN, FP, and FN
View table
Table 1. Definitions of TP, TN, FP, and FN
Real category Predictive value Definition
1 1 $T P$
1 0 $F P$
0 1 $T N$
0 0 $F N$

Table 2. Experimental results of different backbone networks

View table

Table 2. Experimental results of different backbone networks

Algorithm framework	Backbone network	AP50		P		R		mAP
Algorithm framework	Backbone network	Normal	Defect	Normal	Defect	Normal	Defect	mAP
CenterNet	ResNet18	80.80	14.95	96.86	50.00	41.48	0.79	47.87
	ResNet50	94.72	76.46	94.20	89.47	84.08	53.54	85.59
	ResNet101	91.65	62.70	98.86	84.00	77.80	33.07	77.17
	DLANet34	93.61	62.61	99.41	81.40	75.56	27.56	78.11
	Res2Net50	94.26	82.29	95.93	91.09	84.53	72.44	88.27

Table 3. Ablation experiment
View table
Table 3. Ablation experiment
No. Res2Net50 ASPP ECA SE P R
Normal Defect Normal Defect
1 √ 98.68 92.22 83.63 65.35
2 √ √ 97.16 92.52 84.53 77.95
3 √ √ √ 95.73 94.50 85.43 81.10
4 √ √ √ √ 97.26 96.36 87.44 83.46

Table 4. Ablation experiment
View table
Table 4. Ablation experiment
No. Res2Net50 ASPP ECA SE AP50 F1 mAP /%
Normal /% Defect /% Normal Defect
1 √ 93.29 86.24 0.91 0.76 89.77
2 √ √ 95.33 88.58 0.90 0.85 91.96
3 √ √ √ 94.73 92.27 0.90 0.87 93.50
4 √ √ √ √ 95.88 94.81 0.92 0.89 95.35

Table 5. Comparison experiment of adding different attention on two branches
View table
Table 5. Comparison experiment of adding different attention on two branches
The first branch The second branch AP50 P R mAP
Normal Defect Normal Defect Normal Defect
ECA ECA 95.28 89.70 97.74 93.33 87.44 77.17 92.49
ECA SE 95.88 94.81 97.26 96.36 87.44 83.46 95.35

Table 6. Comparison of different algorithms

View table

Table 6. Comparison of different algorithms

Algorithm framework （backbone）	AP50		P		R		mAP	FPS
Algorithm framework （backbone）	Normal	Defect	Normal	Defect	Normal	Defect	mAP	FPS
SSD（VGG16）	91.02	88.87	93.75	92.86	87.5	88.67	89.95	58.62
RetinaNet（ResNet50）	88.70	86.38	95.91	88.03	78.92	84.1	87.54	44.30
FasterRCNN（VGG16）	96.25	70.49	73.47	42.91	96.86	95.28	83.37	29.16
YOLOv3（Darknet53）	93.90	87.45	91.61	91.27	93.05	90.55	90.68	75.63
CenterNet（ResNet50）	94.72	76.46	94.20	89.47	84.08	53.54	85.59	97.51
Proposed algorithm	95.88	94.81	97.26	96.36	87.44	83.46	95.35	65.95

Tools

Get Citation

Copy Citation Text

Lirong Li, Peng Chen, Yunliang Zhang, Kai Zhang, Wei Xiong, Pengcheng Gong. Insulator Defect Detection Based on Multi-Scale Feature Coding and Dual Attention Fusion[J]. Laser & Optoelectronics Progress, 2022, 59(24): 2410008

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites