Method for Measuring Visibility on Foggy Highways Based on Depth Estimation of Encoder and Decoder

Peng Peng; Yucheng Dong; Jiachun Li; Yitao Yao

doi:10.3788/AOS250467

Acta Optica Sinica, Volume. 45, Issue 8, 0810002(2025)

Method for Measuring Visibility on Foggy Highways Based on Depth Estimation of Encoder and Decoder

Peng Peng^1,3、**, Yucheng Dong¹, Jiachun Li^2、*, and Yitao Yao¹

¹School of Electrical and Control Engineering, Shaanxi University of Science & Technology, Xi’an 710021, Shaanxi , China

²Highway School, Chang’an University, Xi’an 710064, Shaanxi , China

³Institute of Flexible Electronics, Northwestern Polytechnical University, Xi’an 710129, Shaanxi , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(16)

Fig. 1. Framework of foggy visibility estimation algorithm

Download full size

Fig. 2. Image segmentation results. (a) Original images; (b) bright area masks; (c) segmentation results

Download full size

Fig. 3. Video image segmentation. (a) Video image; (b) segmentation result

Download full size

Fig. 4. Image of result of local entropy method

Download full size

Fig. 5. Acquisition of transmission maps. (a) Video image; (b) dark channel image; (c) original transmission map; (d) optimized transmission map

Download full size

Fig. 6. Encoder-decoder network based on DenseNet-169

Download full size

Fig. 7. Mechanism of dense concat

Download full size

Fig. 8. Improvement of Dense Block. (a) Dense Block; (b) Dense Block-B

Download full size

Fig. 9. Structure of CBAM

Download full size

Fig. 10. Visibility monitoring station. (a) Device deployment; (b) cloud platform monitoring

Download full size

Fig. 11. Surveillance video frames captured at 5-minute interval. (a) 14:25; (b) 14:30; (c) 14:35; (d) 14:40; (e) 14:45; (f) 14:50; (g) 14:55; (h) 15:00; (i) 15:05; (j) 15:10; (k) 15:15; (l) 15:20

Download full size

Fig. 12. Comparison of results based on different model methods. (a) Visibility curves; (b) error curves

Download full size

Table 1. Comparison of performance of different depth estimation network algorithms

View table

Table 1. Comparison of performance of different depth estimation network algorithms

Model	A_R	RMSE /m	A/%
Model	A_R	RMSE /m	$δ$ is <1.25	$δ$ is <1.25²	$δ$ is <1.25³
BTS	0.116	0.390	81.0	84.1	86.7
LapDepth	0.101	0.432	84.1	84.8	85.6
MonoDepth	0.109	0.440	85.4	85.3	84.1
CADepth	0.103	0.408	85.7	86.0	85.1
Lite-Mono	0.099	0.389	87.9	86.5	85.5
DenseNet-169	0.097	0.385	88.6	86.5	85.4

Table 2. Comparison of network performance under different attention mechanisms

View table

Table 2. Comparison of network performance under different attention mechanisms

Model	A_R	RMSE /m	A/%
Model	A_R	RMSE /m	$δ$ is <1.25	$δ$ is <1.25²	$δ$ is <1.25³
DenseNet-169	0.097	0.385	88.6	86.5	85.4
Dense+CA	0.088	0.323	87.4	90.1	89.2
Dense+ECA	0.098	0.379	92.9	86.8	87.3
Dense+SSA	0.081	0.310	86.3	88.9	90.3
Dense+CBAM	0.082	0.305	89.6	90.4	89.5

Table 3. Levels of visibility
View table
Table 3. Levels of visibility
Level Qualitative descriptor Visibility
1 Excellent V≥10 km
2 Good 2 km≤V<10 km
3 Average 1 km≤V<2 km
4 Poor 500 m≤V<1 km
5 Bad 50 m≤V<500 m
6 Terrible V<50 m

Table 4. Visibility estimation results of proposed method

View table

Table 4. Visibility estimation results of proposed method

Time	Figure	Visibility detector		Proposed method		Error
Time	Figure	Visibility /m	Level	Visibility /m	Level	Error
14：25	Fig. 11（a）	632	4	573	4	59
14：30	Fig. 11（b）	697	4	615	4	82
14：35	Fig. 11（c）	595	4	646	4	51
14：40	Fig. 11（d）	681	4	716	4	35
14：45	Fig. 11（e）	704	4	629	4	75
14：50	Fig. 11（f）	670	4	587	4	83
14：55	Fig. 11（g）	688	4	615	4	73
15：00	Fig. 11（h）	512	4	459	5	53
15：05	Fig. 11（i）	793	4	683	4	110
15：10	Fig. 11（j）	653	4	721	4	68
15：15	Fig. 11（k）	834	4	736	4	98
15：20	Fig. 11（l）	860	4	771	4	89

Tools

Get Citation

Copy Citation Text

Peng Peng, Yucheng Dong, Jiachun Li, Yitao Yao. Method for Measuring Visibility on Foggy Highways Based on Depth Estimation of Encoder and Decoder[J]. Acta Optica Sinica, 2025, 45(8): 0810002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites