Improved Flame Detection Algorithm Based on Salient Target Detection

Ming Lu; Jingang Tan; Zhiyi Zhang; Ming Chen; Wei He

doi:10.3788/LOP202259.0410012

Laser & Optoelectronics Progress, Volume. 59, Issue 4, 0410012(2022)

Improved Flame Detection Algorithm Based on Salient Target Detection

Ming Lu^1,2, Jingang Tan^1,2, Zhiyi Zhang¹, Ming Chen³, and Wei He^1、*

¹Key Laboratory of Wireless Sensor Networks and Communication, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 201800, China

²University of Chinese Academy of Sciences, Beijing 100864, China

³Wuxi Hi-Tech Nano SensoringNet R&D Center of Chinese Academy of Sciences, Wuxi, Jiangsu 214135, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(15)

Fig. 1. Flow chart of proposed algorithm

Download full size

Fig. 2. Improved fire detection model

Download full size

Fig. 3. Feature extraction network

Download full size

Fig. 4. Short connection branch network

Download full size

Fig. 5. Dilated convolution module

Download full size

Fig. 6. Attention mechanism network

Download full size

Fig. 7. Bi-FPN

Download full size

Fig. 8. Part of sample dataset. (a)-(c) Positive samples; (d)-(f) negative samples

Download full size

Fig. 9. Partial test results

Download full size

Table 1. Performance analysis of feature extraction networks with different depth
View table
Table 1. Performance analysis of feature extraction networks with different depth
Parameter ResNet-18 ResNet-34 ResNet-50
MAE 0.0101 0.0138 0.0104
F_β_fg 0.8812 0.8815 0.8943
F_β_bg 0.9010 0.8802 0.8734

Table 2. Performance comparison between proposed algorithm and YOLOv4, RetinaNet
View table
Table 2. Performance comparison between proposed algorithm and YOLOv4, RetinaNet
Parameter Proposed algorithm YOLOv4 RetinaNet
Precision 0.890 0.900 0.802
Recall 0.880 0.880 0.773
F_β 0.884 0.889 0.788

Table 3. Performance comparison between proposed algorithm and BASNet, PICANet
View table
Table 3. Performance comparison between proposed algorithm and BASNet, PICANet
Parameter Proposed algorithm BASNet PICANet
F_β 0.881 0.877 0.861
MAE 0.010 0.013 0.023

Table 4. Comparison of reasoning speed and model size of different algorithms
View table
Table 4. Comparison of reasoning speed and model size of different algorithms
Parameter Proposed algorithm YOLOv4 RetinaNet BASNet PICANet
Detection speed /（frame⋅s^-1） 4 4 4.5 2 3
Model size /MB 121.5 256 145.7 348.5 188.9

Table 5. Comparison of detection results between proposed algorithm and YOLOv4
View table
Table 5. Comparison of detection results between proposed algorithm and YOLOv4
Parameter Proposed algorithm YOLOv4
$N_{F P}$ 0.020 0.200
$N_{F N}$ 0.003 0.067
$N_{T P}$ 0.980 0.800
$N_{T N}$ 0.997 0.933

Table 6. Ablation experiments on flame dataset
View table
Table 6. Ablation experiments on flame dataset
Single ResNet Double Bi-FPN Dilated Conv Attention mechanism MAE F_β
√ 0.085 0.8079
√ √ 0.043 0.8521
√ √ √ 0.032 0.8658
√ √ √ √ 0.010 0.8840
√ √ √ √ 0.032 0.8498

Tools

Get Citation

Copy Citation Text

Ming Lu, Jingang Tan, Zhiyi Zhang, Ming Chen, Wei He. Improved Flame Detection Algorithm Based on Salient Target Detection[J]. Laser & Optoelectronics Progress, 2022, 59(4): 0410012

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites