An Efficient Algorithm for Extracting Railway Tracks Based on Spatial-Channel Graph Convolutional Network and Deep Neural Residual Network
Abstract
:1. Introduction
- Establishment of a segmentation dataset: High-resolution images of train stations in southern China were obtained via drone photography, annotated, calibrated, and manually corrected using Labelme 3.16.7. Data augmentation techniques such as cutting, rotating, and flipping were employed to expand the dataset, resulting in a railway dataset comprising 13,285 images.
- Introduction of deep neural residual network and graph convolutional network: The deep neural residual network consists of multiple residual structures in series and parallel, combined with depthwise separable convolutions and inverted residual structures, enabling the sharing of multi-scale convolution kernel parameter information and addressing the issue of train carriage occlusion. The graph convolution structure separately processes spatial and channel features, enhancing the representation of track features.
- Optimization of the overall structure: The original RELU activation function was replaced with PRELU, and the number of model layers was reduced to ensure segmentation efficiency. The scSE attention mechanism module was added to the encoder and decoder to suppress unimportant features, reduce noise interference, and enhance foreground response, thereby improving segmentation accuracy and robustness.
2. Materials and Methods
2.1. Algorithm Flow
2.2. Improved U-Net Network Structure
2.3. Deep Neural Residual Network
2.4. Spatial Channel Graph Convolutional Network
2.5. scSE Module
3. Experimental Data and Evaluation Indexes
3.1. Experimental Data
3.2. Experimental Environment
3.3. Evaluation Index
3.4. Experimental Analysis
3.4.1. Visual Analysis of Loss Function
3.4.2. Impact of the scSE on Model Performance
3.4.3. Comparative Experimental Analysis
3.4.4. Analysis of the Ablation Experiment
4. Discussion
5. Conclusions
Author Contributions
Funding
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- Shvetsov, A.V.; Dronichev, A.V.; Kuzmina, N.A.; Shvetsova, S.V. Analysis of the Directions of Optimization of the Process of Ensuring Transportation Security in Railway Transport. Transp. Res. Procedia 2023, 68, 579–584. [Google Scholar] [CrossRef]
- Gholamizadeh, K.; Zarei, E.; Yazdi, M. Railway Transport and Its Role in the Supply Chains: Overview, Concerns, and Future Direction. In The Palgrave Handbook of Supply Chain Management; Sarkis, J., Ed.; Springer International Publishing: Cham, Switzerland, 2024; pp. 769–796. ISBN 978-3-031-19883-0. [Google Scholar]
- Song, T.; Schonfeld, P.; Pu, H. A Review of Alignment Optimization Research for Roads, Railways and Rail Transit Lines. IEEE Trans. Intell. Transp. Syst. 2023, 24, 4738–4757. [Google Scholar] [CrossRef]
- Chen, S.-B.; Ji, Y.-X.; Tang, J.; Luo, B.; Wang, W.-Q.; Lv, K. DBRANet: Road Extraction by Dual-Branch Encoder and Regional Attention Decoder. IEEE Geosci. Remote Sens. Lett. 2021, 19, 3002905. [Google Scholar] [CrossRef]
- Mo, S.; Shi, Y.; Yuan, Q.; Li, M. A Survey of Deep Learning Road Extraction Algorithms Using High-Resolution Remote Sensing Images. Sensors 2024, 24, 1708. [Google Scholar] [CrossRef] [PubMed]
- Chen, H.; Li, Z.; Wu, J.; Xiong, W.; Du, C. SemiRoadExNet: A Semi-Supervised Network for Road Extraction from Remote Sensing Imagery via Adversarial Learning. ISPRS J. Photogramm. Remote Sens. 2023, 198, 169–183. [Google Scholar] [CrossRef]
- Xu, H.; He, H.; Zhang, Y.; Ma, L.; Li, J. A Comparative Study of Loss Functions for Road Segmentation in Remotely Sensed Road Datasets. Int. J. Appl. Earth Obs. Geoinf. 2023, 116, 103159. [Google Scholar] [CrossRef]
- Khan, N.; Shahid, Z.; Alam, M.M.; Sajak, A.A.B.; Nazar, M.; Mazliham, M.S. A Novel Deep Learning Technique to Detect Electricity Theft in Smart Grids Using AlexNet. IET Renew. Power Gener. 2024, 18, 941–958. [Google Scholar] [CrossRef]
- Qian, L.; Huang, H.; Xia, X.; Li, Y.; Zhou, X. Automatic Segmentation Method Using FCN with Multi-Scale Dilated Convolution for Medical Ultrasound Image. Vis. Comput. 2023, 39, 5953–5969. [Google Scholar] [CrossRef]
- Ansari, M.Y.; Yang, Y.; Meher, P.K.; Dakua, S.P. Dense-PSP-UNet: A Neural Network for Fast Inference Liver Ultrasound Segmentation. Comput. Biol. Med. 2023, 153, 106478. [Google Scholar] [CrossRef]
- Jabbar, A.; Naseem, S.; Mahmood, T.; Saba, T.; Alamri, F.S.; Rehman, A. Brain Tumor Detection and Multi-Grade Segmentation through Hybrid Caps-VGGNet Model. IEEE Access 2023, 11, 72518–72536. [Google Scholar] [CrossRef]
- Wang, J.; Song, J.; Chen, M.; Yang, Z. Road Network Extraction: A Neural-Dynamic Framework Based on Deep Learning and a Finite State Machine. Int. J. Remote Sens. 2015, 36, 3144–3169. [Google Scholar] [CrossRef]
- Zhou, L.; Zhang, C.; Wu, M. D-LinkNet: LinkNet with Pretrained Encoder and Dilated Convolution for High Resolution Satellite Imagery Road Extraction. In Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Salt Lake City, UT, USA, 18–22 June 2018; pp. 182–186. [Google Scholar]
- Kipf, T.N.; Welling, M. Semi-Supervised Classification with Graph Convolutional Networks. arXiv 2016, arXiv:1609.02907. [Google Scholar]
- Lu, Y.; Chen, Y.; Zhao, D.; Chen, J. Graph-FCN for Image Semantic Segmentation. In Advances in Neural Networks—ISNN 2019; Lu, H., Tang, H., Wang, Z., Eds.; Lecture Notes in Computer Science; Springer International Publishing: Cham, Switzerland, 2019; Volume 11554, pp. 97–105. ISBN 978-3-030-22795-1. [Google Scholar]
- Chen, L.; Zhang, Q. DDGCN: Graph Convolution Network Based on Direction and Distance for Point Cloud Learning. Vis. Comput. 2023, 39, 863–873. [Google Scholar] [CrossRef]
- Yu, H.; Ai, T.; Yang, M.; Huang, L.; Gao, A. Automatic Segmentation of Parallel Drainage Patterns Supported by a Graph Convolution Neural Network. Expert Syst. Appl. 2023, 211, 118639. [Google Scholar] [CrossRef]
- Sandler, M.; Howard, A.; Zhu, M.; Zhmoginov, A.; Chen, L.-C. Mobilenetv2: Inverted Residuals and Linear Bottlenecks. In Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 18–22 June 2018; pp. 4510–4520. [Google Scholar]
- Dai, Y.; Li, C.; Su, X.; Liu, H.; Li, J. Multi-Scale Depthwise Separable Convolution for Semantic Segmentation in Street–Road Scenes. Remote Sens. 2023, 15, 2649. [Google Scholar] [CrossRef]
- Huang, T.; Chen, J.; Jiang, L. DS-UNeXt: Depthwise Separable Convolution Network with Large Convolutional Kernel for Medical Image Segmentation. Signal Image Video Process. 2023, 17, 1775–1783. [Google Scholar] [CrossRef]
- Zhang, Y.; Xu, S.; Zhang, L.; Jiang, W.; Alam, S.; Xue, D. Short-Term Multi-Step-Ahead Sector-Based Traffic Flow Prediction Based on the Attention-Enhanced Graph Convolutional LSTM Network (AGC-LSTM). Neural Comput. Appl. 2024, 1–20. [Google Scholar] [CrossRef]
- Zhu, J.-Y.; Park, T.; Isola, P.; Efros, A.A. Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks. In Proceedings of the 2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 22–29 October 2017; pp. 2223–2232. [Google Scholar]
- Weng, Y.; Huang, X.; Chen, X.; He, J.; Li, Z.; Yi, H. Research on Railway Track Extraction Method Based on Edge Detection and Attention Mechanism. IEEE Access 2024, 12, 26550–26561. [Google Scholar] [CrossRef]
- Weng, Y.; Li, Z.; Chen, X.; He, J.; Liu, F.; Huang, X.; Yang, H. A Railway Track Extraction Method Based on Improved DeepLabV3+. Electronics 2023, 12, 3500. [Google Scholar] [CrossRef]
- Saeedizadeh, N.; Minaee, S.; Kafieh, R.; Yazdani, S.; Sonka, M. COVID TV-Unet: Segmenting COVID-19 Chest CT Images Using Connectivity Imposed Unet. Comput. Methods Programs Biomed. Update 2021, 1, 100007. [Google Scholar] [CrossRef]
- Zhao, S.; Feng, Z.; Chen, L.; Li, G. DANet: A Semantic Segmentation Network for Remote Sensing of Roads Based on Dual-ASPP Structure. Electronics 2023, 12, 3243. [Google Scholar] [CrossRef]
- Mehta, S.; Rastegari, M. MobileViT: Light-Weight, General-Purpose, and Mobile-Friendly Vision Transformer 2022. arXiv 2022, arXiv:2110.02178. [Google Scholar]
- Zhou, G.; Chen, W.; Gui, Q.; Li, X.; Wang, L. Split Depth-Wise Separable Graph-Convolution Network for Road Extraction in Complex Environments from High-Resolution Remote-Sensing Images. IEEE Trans. Geosci. Remote Sens. 2021, 60, 5614115. [Google Scholar] [CrossRef]
- Li, X.; Yang, Y.; Zhao, Q.; Shen, T.; Lin, Z.; Liu, H. Spatial Pyramid Based Graph Reasoning for Semantic Segmentation. In Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 13–19 June 2020; pp. 8950–8959. [Google Scholar]
- Yu, D.; Yang, Y.; Zhang, R.; Wu, Y. Knowledge Embedding Based Graph Convolutional Network. In Proceedings of the WWW ‘21: The Web Conference 2021, Ljubljana, Slovenia, 19 April 2021; pp. 1619–1628. [Google Scholar]
- Balochian, S.; Baloochian, H. Edge Detection on Noisy Images Using Prewitt Operator and Fractional Order Differentiation. Multimed. Tools Appl. 2022, 81, 9759–9770. [Google Scholar] [CrossRef]
- Zhu, W.; Li, H.; Cheng, X.; Jiang, Y. A Multi-Task Road Feature Extraction Network with Grouped Convolution and Attention Mechanisms. Sensors 2023, 23, 8182. [Google Scholar] [CrossRef]
- Roy, A.G.; Navab, N.; Wachinger, C. Concurrent Spatial and Channel ‘Squeeze & Excitation’ in Fully Convolutional Networks. In Medical Image Computing and Computer Assisted Intervention—MICCAI 2018; Frangi, A.F., Schnabel, J.A., Davatzikos, C., Alberola-López, C., Fichtinger, G., Eds.; Lecture Notes in Computer Science; Springer International Publishing: Cham, Switzerland, 2018; Volume 11070, pp. 421–429. ISBN 978-3-030-00927-4. [Google Scholar]
Position | Railway | DeepGlobe | ||||
---|---|---|---|---|---|---|
MIoU | OA | Recall | MIoU | OA | Recall | |
None | 87.64% | 96.25% | 85.12% | 66.54% | 95.71% | 79.65% |
Only Encoder | 87.23% | 96.02% | 87.14% | 72.17% | 95.94% | 79.76% |
Only Decoder | 87.58% | 97.37% | 86.78% | 72.39% | 95.83% | 79.88% |
Both | 88.47% | 97.62% | 88.90% | 75.81% | 96.02% | 80.28% |
Aggregation Strategies | Railway | DeepGlobe | ||||
---|---|---|---|---|---|---|
MIoU | OA | Recall | MIoU | OA | Recall | |
Max-out | 87.20% | 95.56% | 86.35% | 77.41% | 94.72% | 79.43% |
Addition | 86.39% | 95.08% | 83.63% | 77.63% | 92.97% | 77.45% |
Multiplication | 82.12% | 94.10% | 70.22% | 70.61% | 87.42% | 75.34% |
Concatenation | 83.91% | 93.72% | 82.79% | 76.23% | 92.35% | 78.29% |
Weighted fusion | 88.72% | 96.49% | 86.68% | 78.49% | 95.16% | 80.26% |
Original | PSPNet | UNet++ | U-Net | Deeplab V3+ | Proposed Method | Ground Truth |
---|---|---|---|---|---|---|
Model | Rail IoU | MIoU | OA | F1 Score | Recall | Precision |
---|---|---|---|---|---|---|
PSPNet | 77.89% | 86.43% | 96.24% | 87.59% | 87.65% | 85.54% |
U-Net | 77.54% | 86.24% | 96.43% | 87.35% | 86.62% | 85.19% |
UNet++ | 76.72% | 85.73% | 95.67% | 84.48% | 87.71% | 82.93% |
Deeplab V3+ | 76.55% | 86.03% | 97.08% | 87.14% | 88.45% | 85.88% |
Propose Method | 79.49% | 87.98% | 97.48% | 88.18% | 89.79% | 88.57% |
Original | PSPNet | UNet++ | U-Net | Deeplab V3+ | Proposed Method | Ground Truth |
---|---|---|---|---|---|---|
Model | Road IoU | MIoU | OA | F1 Score | Recall | Precision |
---|---|---|---|---|---|---|
PSPNet | 73.38% | 69.80% | 96.08% | 77.90% | 64.63% | 76.56% |
U-Net | 73.07% | 70.02% | 97.10% | 70.73% | 76.22% | 77.28% |
UNet++ | 73.49% | 75.00% | 96.72% | 78.11% | 76.79% | 75.03% |
Deeplab V3+ | 78.19% | 72.31% | 96.49% | 73.98% | 66.68% | 77.05% |
Propose Method | 77.96% | 77.33% | 96.98% | 79.00% | 82.38% | 78.13% |
GCN | Residual | scSE | MIoU | F1 Score | Recall | Precision |
---|---|---|---|---|---|---|
√ | × | × | 87.93% | 87.85% | 89.42% | 84.33% |
× | √ | × | 86.90% | 86.92% | 89.40% | 82.05% |
√ | √ | × | 88.20% | 88.46% | 88.59% | 85.35% |
× | √ | √ | 88.36% | 88.58% | 89.44% | 88.02% |
√ | × | √ | 88.05% | 86.98% | 89.96% | 84.59% |
√ | √ | √ | 88.23% | 89.25% | 90.79% | 89.62% |
GCN | Residual | scSE | MIoU | F1 Score | Recall | Precision |
---|---|---|---|---|---|---|
√ | × | × | 69.81% | 76.77% | 71.03% | 76.23% |
× | √ | × | 63.91% | 77.25% | 79.27% | 76.09% |
√ | √ | × | 67.01% | 79.36% | 79.85% | 75.72% |
× | √ | √ | 67.03% | 79.84% | 73.67% | 77.66% |
√ | × | √ | 70.00% | 79.32% | 73.26% | 79.84% |
√ | √ | √ | 76.45% | 81.93% | 81.86% | 80.13% |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Published by MDPI on behalf of the International Society for Photogrammetry and Remote Sensing. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Weng, Y.; Xu, M.; Chen, X.; Peng, C.; Xiang, H.; Xie, P.; Yin, H. An Efficient Algorithm for Extracting Railway Tracks Based on Spatial-Channel Graph Convolutional Network and Deep Neural Residual Network. ISPRS Int. J. Geo-Inf. 2024, 13, 309. https://doi.org/10.3390/ijgi13090309
Weng Y, Xu M, Chen X, Peng C, Xiang H, Xie P, Yin H. An Efficient Algorithm for Extracting Railway Tracks Based on Spatial-Channel Graph Convolutional Network and Deep Neural Residual Network. ISPRS International Journal of Geo-Information. 2024; 13(9):309. https://doi.org/10.3390/ijgi13090309
Chicago/Turabian StyleWeng, Yanbin, Meng Xu, Xiahu Chen, Cheng Peng, Hui Xiang, Peixin Xie, and Hua Yin. 2024. "An Efficient Algorithm for Extracting Railway Tracks Based on Spatial-Channel Graph Convolutional Network and Deep Neural Residual Network" ISPRS International Journal of Geo-Information 13, no. 9: 309. https://doi.org/10.3390/ijgi13090309