改进U-Net的水岸分割算法
Improve U-Net for Watershore Segmentation
摘要: 我国水域广阔、多样等特点导致难以使用人工的方式实现全面高效地监测、管理，使用现代信息技术辅助水域管理可以极大地提高效率。水岸分割是水域管理信息化的关键技术，传统的水岸分割算法依赖手工特征提取需要针对特定的环境调整参数，适应性不高不适用于多变的天气条件和复杂的河岸场景。基于学习的方法比传统方法更能适应不同的环境，本文使用基于深度学习的U-Net模型进行水岸分割，并引入空间注意力模块调整不同空间的特征权重，使用图像处理方法模拟环境中的雨雾进行数据增强。实验结果表明模型分割MIoU达97.60%，较已有的算法性能有显著提升。所提模型在极端挑战环境中仍能保持较高的分割准确率，与传统算法相比具有优秀的环境适应性。
Abstract: The characteristics of the vast and diverse waters in the country make it difficult to realize efficient monitoring and management manually. Modern information technology has potential for improving the efficiency of water management greatly. Watershore segmentation is the key technology for water management informatization. Traditional watershore segmentation algorithms rely on manual feature extraction and need to adjust parameters for specific environments, which can’t be used in watershore segmentation in a variety of weather conditions and in complex inland river scenarios due to its poor environment adaptability. Compared with traditional methods, learning-based methods are more adaptable to different environments. In this paper, U-Net is used for the water bank segmentation, which is a deep learning method. The spatial attention module is introduced to adjust the feature weights of different spaces, and the image processing method is used to simulate the rain and fog in the environment for data enhancement. The experimental results show that the model segmentation MIoU is 97.60%, which is significantly improved compared with the existing algorithm. The proposed model maintains high accuracy even in extremely challenging environments, having the property of impressive environmental adaptability compared with the traditional algorithm.
文章引用：姚福飞, 柏利志, 周田. 改进U-Net的水岸分割算法[J]. 计算机科学与应用, 2022, 12(12): 2875-2883. https://doi.org/10.12677/CSA.2022.1212292

参考文献

[1] 宋庆辉, 杨志峰. 对我国城市河流综合管理的思考[J]. 水科学进展, 2002, 13(3): 377-382.
[2] 李晓玉, 王鹏, 邵光艺, 等. 基于水面无人船快速监测的温榆河-北运河水质分析与评价[J]. 环境化学, 2022, 41(5): 1568-1578.
[3] 鲍江, 陶青川, 张鹏. 基于图像处理的水位线检测算法[J]. 水电能源科学, 2015, 33(4): 96-99,210.
[4] 沈建军, 陶青川, 肖卓. 结合改进Deeplab v3+网络的水岸线检测算法[J]. 中国图象图形学报, 2019, 24(12): 2174-2182.
[5] Fefilatyev, S. and Goldgof, D. (2008) Detection and Tracking of Marine Vehicles in Video. 2008 IEEE 19th International Conference on Pattern Recognition, Tampa, 8-11 December 2008, 1-4. [Google Scholar] [CrossRef] 
[6] 董宇星, 刘伟宁. 基于灰度特性的海天背景小目标检测[J]. 中国光学与应用光学, 2010, 3(3): 252-256.
[7] 郑又能, 肖宇峰, 寸超, 等. 基于改进区域生长的水岸线提取方法研究[J]. 计算机应用研究, 2020, 37(6): 1876-1881.
[8] 彭明阳, 王建华, 闻祥鑫, 丛晓奕. 结合HSV空间的水面图像特征水岸线检测[J]. 中国图象图形学报, 2018, 23(4): 526-533.
[9] Krizhevsky, A., Sutskever, I. and Hinton, G.E. (2017) Imagenet Classification with Deep Convolutional Neural Networks. Communications of the ACM, 60, 84-90. [Google Scholar] [CrossRef] 
[10] Long, J., Shelhamer, E. and Darrell, T. (2015) Fully Convolutional Networks for Semantic Segmentation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recogni-tion, Boston, 7-12 June 2015, 3431-3440. [Google Scholar] [CrossRef] 
[11] Ronneberger, O., Fischer, P. and Brox, T. (2015) U-Net: Con-volutional Networks for Biomedical Image Segmentation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention, Springer, Berlin, 234-241. [Google Scholar] [CrossRef] 
[12] Chen, L.-C., Zhu, Y., Papandreou, G., et al. (2018) Encod-er-Decoder with Atrous Separable Convolution for Semantic Image Segmentation. Proceedings of the European Confer-ence on Computer Vision (ECCV), Munich, 8-14 September 2018, 801-818. [Google Scholar] [CrossRef] 
[13] Zhao, H., Shi, J., Qi, X., et al. (2017) Pyramid Scene Parsing Network. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, 21-26 July 2017, 2881-2890. [Google Scholar] [CrossRef] 
[14] Yin, Y., Guo, Y., Deng, L. and Chai, B. (2022) Improved PSPNet-Based Water Shoreline Detection in Complex Inland River Scenarios. Complex & Intelligent Systems, 1-13. [Google Scholar] [CrossRef] 
[15] Erdem, F., Bayram, B., Bakirman, T., et al. (2021) An En-semble Deep Learning Based Shoreline Segmentation Approach (WaterNet) from Landsat 8 OLI Images. Advances in Space Research, 67, 964-974. [Google Scholar] [CrossRef] 
[16] von Braun, M.-S., Frenzel, P., Kading, C. and Fuchs, M. (2020) Utilizing Mask R-CNN for Waterline Detection in CANOE Sprint Video Analysis. Proceedings of the IEEE/CVF Con-ference on Computer Vision and Pattern Recognition Workshops, Seattle, 14-19 June 2020, 876-877. [Google Scholar] [CrossRef] 
[17] McLeay, A.J., McGhie, A., Briscoe, D., et al. (2021) Deep Convolutional Neural Networks with Transfer Learning for Waterline Detection in Mussel Farms. 2021 IEEE Symposium Series on Computational Intelligence (SSCI), Orlando, 5-7 December 2021, 1-8. [Google Scholar] [CrossRef] 
[18] Jaderberg, M., Simonyan, K. and Zisserman, A. (2015) Spatial Transformer Networks. Proceedings of the 28th International Conference on Neural Information Processing Systems, Volume 2, 2017-2025.
[19] Hu, J., Shen, L. and Sun, G. (2018) Squeeze-and-Excitation Networks. Proceed-ings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, 18-22 June 2018, 7132-7141. [Google Scholar] [CrossRef] 
[20] Bovcon, B., Muhovič, J., Perš, J. and Kristan, M. (2019) The MaSTr1325 Dataset for Training Deep USV Obstacle Detection Models. 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Macau, 3-8 November 2019, 3431-3438. [Google Scholar] [CrossRef] 
[21] Cheng, Y., Jiang, M., Zhu, J. and Liu, Y. (2021) Are We Ready for Unmanned Surface Vehicles in Inland Waterways? The USVInland Multisensor Dataset and Benchmark. IEEE Robotics and Automation Letters, 6, 3964-3970. [Google Scholar] [CrossRef]