TY - JOUR
T1 - The enhancement of catenary image with low visibility based on multi-feature fusion network in railway industry
AU - Chen, Yuwen
AU - Song, Bin
AU - Du, Xiaojiang
AU - Guizani, Nadra
N1 - Publisher Copyright:
© 2020 Elsevier B.V.
PY - 2020/2/15
Y1 - 2020/2/15
N2 - In the Industrial Internet of Things (IIoT), the security and efficiency are indispensable. For the railway industry, the video from inspection vehicle would be influenced by various factors with low visibility and hard for high level vision task, such as the fault diagnosis of catenary system. In this paper, we propose a method based on the multi-feature fusion network to improve the quality and visual effect of the catenary images. The transmission map is learned from the multi-scale and multi-feature fusion network, which would learn coarse and fine details and combine the latent features. In the catenary image, the sky and non-sky regions are segmented through multiple accommodative thresholds to estimate the atmospheric light value. With the refinement of transmission map, the restored catenary image is obtained through the atmospheric scattering model. In the experimental results, it can be seen that the proposed method can improve the clarity of catenary image in haze. The quantitative evaluation shows that it has better visual effect compared with the other traditional methods.
AB - In the Industrial Internet of Things (IIoT), the security and efficiency are indispensable. For the railway industry, the video from inspection vehicle would be influenced by various factors with low visibility and hard for high level vision task, such as the fault diagnosis of catenary system. In this paper, we propose a method based on the multi-feature fusion network to improve the quality and visual effect of the catenary images. The transmission map is learned from the multi-scale and multi-feature fusion network, which would learn coarse and fine details and combine the latent features. In the catenary image, the sky and non-sky regions are segmented through multiple accommodative thresholds to estimate the atmospheric light value. With the refinement of transmission map, the restored catenary image is obtained through the atmospheric scattering model. In the experimental results, it can be seen that the proposed method can improve the clarity of catenary image in haze. The quantitative evaluation shows that it has better visual effect compared with the other traditional methods.
KW - Catenary system image
KW - Deep learning
KW - Low visibility
UR - http://www.scopus.com/inward/record.url?scp=85078431314&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85078431314&partnerID=8YFLogxK
U2 - 10.1016/j.comcom.2020.01.040
DO - 10.1016/j.comcom.2020.01.040
M3 - Article
AN - SCOPUS:85078431314
SN - 0140-3664
VL - 152
SP - 200
EP - 205
JO - Computer Communications
JF - Computer Communications
ER -