Desirable energy space identification of clean and self-consistent energy along railways
-
摘要: 为了智能识别铁路沿线的清洁自洽能源宜能空间，构建了一个包含210张分辨率为4 800像素× 2 986像素的铁路遥感图像数据集；针对遥感图像处理领域中多尺度融合单元通道信息不兼容的问题，提出了自注意力机制模块，以增强对多尺度特征的捕捉能力；针对不同尺度高分辨率遥感图像预测结果存在的差异性，提出了多尺度一致性正则化方法(MSCR)，以增强模型对不同尺度图像处理的鲁棒性；综合自注意力机制模块与多尺度一致性正则化方法，对经典的图像分割HRNetV2方法进行了改进，构建了MSCR-HRNetV2方法；在自建的铁路遥感图像数据集和公开的Potsdam遥感图像数据集分别对MSCR-HRNetV2方法进行验证。研究结果表明：在铁路遥感图像数据集上，改进的MSCR-HRNetV2方法取得了81.37%的平均交并比，相较于原HRNetV2方法提高了3.13%，与主流图像分割方法DeepLabV3+相比，提高了3.86%；在Potsdam遥感图像数据集上，MSCR-HRNetV2方法的平均交并比达到了75.96%，相比HRNetV2方法提高了2.01%，与DeepLabV3+相比提高了2.19%。可见，改进的MSCR-HRNetV2方法显著提高了高分辨率遥感图像的语义分割性能，从而智能识别铁路沿线的清洁自洽能源宜能空间，为交通能源融合系统的规划和设计提供了重要的技术支撑。Abstract: In order to intelligently identify the desirable energy space of clean and self-consistent energy along railways, a remote sensing image dataset containing 210 railway images with a resolution of 4 800 pixel×2 986 pixel was constructed. To address the problem of incompatible channel information of multi-scale fusion units in remote sensing image processing, a self-attention mechanism module was proposed to enhance the ability to capture multi-scale features. To address the discrepancy in the prediction results of remote sensing images with high resolution at different scales, a multi-scale consistency regularization (MSCR) method was proposed to enhance the robustness of the model during image processing. The MSCR-HRNetV2 method was constructed by combining the self-attention mechanism module and the MSCR method to improve the classical image segmentation HRNetV2 method. The MSCR-HRNetV2 method was verified on the self-built remote sensing image dataset of railways and the publicly available Potsdam remote sensing image dataset, respectively. Analysis results show that on the remote sensing image dataset of railways, the improved MSCR-HRNetV2 method achieves a mean intersection over union (MIoU) of 81.37%, which is an improvement of 3.13% compared with the original HRNetV2 method and an improvement of 3.86% compared with the mainstream image segmentation method DeepLabV3+. On the Potsdam remote sensing image dataset, the MIoU of the MSCR-HRNetV2 method reaches 75.96%, which is improved by 2.01% compared to HRNetV2 and 2.19% compared to DeepLabV3+. It can be seen that the improved MSCR-HRNetV2 method significantly improves the semantic segmentation performance of remote sensing images with high resolution, thus intelligently identifying the desirable energy space for clean and self-consistent energy along railways and providing important technical support for the planning and design of the integration system of transportation and energy.
-
图 6 不同方法在数据集1上的语义分割结果
Figure 6. Semantic segmentation results of different methods on dataset 1
图 7 不同方法在数据集2上的语义分割结果
Figure 7. Semantic segmentation results of different methods on dataset 2
表 1 消融试验结果MIoU
Table 1. MIoUs of ablation experiment results
% 方法 SE MSCR 数据集1 数据集2 HRNetV2 78.24 73.33 HRNetV2_A √ 79.88 74.69 HRNetV2_B √ 80.49 74.61 MSCR-HRNetV2 √ √ 81.37 75.96 
下载: 导出CSV
表 2 数据集1中各类别目标MIoU对比
Table 2. Target MIoU comparison of each category on dataset 1
% 类别 HRNetV2 MSCR-HRNetV2 铁路线 78.08 82.41 铁路边坡 77.90 81.69 房屋 68.86 72.88 植被 86.46 88.90 荒地 93.53 94.01
下载: 导出CSV
表 3 数据集2中各类别目标MIoU对比
Table 3. Target MIoU comparison of each category on dataset 2
% 类别 HRNetV2 MSCR-HRNetV2 汽车 77.52 81.56 树木 71.51 72.97 低矮植被 71.97 73.66 建筑 89.26 90.18 不透水表面 49.70 55.84
下载: 导出CSV
表 4 MSCR-HRNetV2与其他方法比较结果
Table 4. Comparison results of MSCR-HRNetV2 and other methods
% 方法 MIoU MPA 数据集1 数据集2 数据集1 数据集2 UNet 74.88 72.06 85.65 82.72 DeepLabV3+ 77.51 71.05 88.24 81.30 PSPNet 72.24 68.76 84.73 79.24 HRNetV2 78.24 73.33 89.21 83.29 MSCR-HRNetV2 81.37 75.96 90.43 85.30
下载: 导出CSV
表 5 不同方法的参数数量和GFLOPs的比较
Table 5. Comparison of params and GFLOPs of different methods
方法 参数数量/106 GFLOPs UNet 24.891 397.095 DeepLabV3+ 5.814 46.494 PSPNet 2.376 5.301 HRNetV2 9.638 32.831 MSCR-HRNetV2 9.750 36.593
下载: 导出CSV
-
[1] DALALA Z, AL-OMARI M, AL-ADDOUS M, et al. Increased renewable energy penetration in national electrical grids constraints and solutions[J]. Energy, 2022, 246: 123361. doi: 10.1016/j.energy.2022.123361 [2] 艾国乐, 郝小礼, 刘仙萍, 等. 高速铁路上空安装光伏系统的节能潜力研究[J]. 太阳能学报, 2023, 44(2): 409-417.AI Guo-le, HAO Xiao-li, LIU Xian-ping, et al. Energy saving potential research of photovoltaic system installed over high-speed railway[J]. Acta Energiae Solaris Sinica, 2023, 44(2): 409-417. (in Chinese) [3] 张舜, 张蜇. 基于光伏发电的铁路与新能源融合潜力评估[J]. 中国铁路, 2023(11): 64-71.ZHANG Shun, ZHANG Zhe. Evaluation of the potential application of new energy in the railway sector based on PV power generation[J]. China Railway, 2023(11): 64-71. (in Chinese) [4] LI Jian, CUI Min. Application of solar PV grid-connected power generation system in Shanghai rail transit[C]//IEEE. 2018 China International Conference on Electricity Distribution. New York: IEEE, 2018: 110-113. [5] ZHONG Zhi-ming, ZHANG Yong-xin, HONG Shen, et al. Optimal planning of distributed photovoltaic generation for the traction power supply system of high-speed railway[J]. Journal of Cleaner Production, 2020, 263: 121394. doi: 10.1016/j.jclepro.2020.121394 [6] NING Fu-wei, JI Li, MA Jing, et al. Research and analysis of a flexible integrated development model of railway system and photovoltaic in China[J]. Renewable Energy, 2021, 175: 853-867. doi: 10.1016/j.renene.2021.04.119 [7] CHEN Zhu-jun, JIANG Ming-kun, QI Ling-fei, et al. Using existing infrastructures of high-speed railways for photovoltaic electricity generation[J]. Resources, Conservation and Recycling, 2022, 178: 106091. doi: 10.1016/j.resconrec.2021.106091 [8] TANG Yu-qi, ZHANG Liang-pei. Urban change analysis with multi-sensor multispectral imagery[J]. Remote Sensing, 2017, 9(3): 252. doi: 10.3390/rs9030252 [9] WU Lin-shan, LU Ming, FANG Le-yuan. Deep covariance alignment for domain adaptive remote sensing image segmentation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-11. [10] PAN Shao-ming, TAO Yu-long, NIE Cong-chong, et al. PEGNet: progressive edge guidance network for semantic segmentation of remote sensing images[J]. IEEE Geoscience and Remote Sensing Letters, 2020(99): 1-5. [11] 常秀红, 李纯斌, 吴静, 等. 基于优化HRNetV2的高分辨率遥感影像土地利用自动分类[J]. 中国土地科学, 2022, 36(2): 96-105.CHANG Xiu-hong, LI Chun-bin, WU Jing, et al. Automatic land use classification based on optimized HRNetV2 High-resolution remote sensing images[J]. China Land Science, 2022, 36(2) : 96-105. (in Chinese) [12] RADMAN A, ZAINAL N, SUANDI S A. Automated segmentation of iris images acquired in an unconstrained environment using HOG-SVM and GrowCut[J]. Digital Signal Processing, 2017, 64: 60-70. doi: 10.1016/j.dsp.2017.02.003 [13] THANH NOI P, KAPPAS M. Comparison of random forest, K-nearest neighbor, and support vector machine classifiers for land cover classification using Sentinel-2 imagery[J]. Sensors, 2017, 18(1): 18. doi: 10.3390/s18010018 [14] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. Imagenet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6): 84-90. doi: 10.1145/3065386 [15] LIN Guo-sheng, MILAN A, SHEN Chun-hua, et al. Refinenet: multi-path refinement networks for high-resolution semantic segmentation[C]//IEEE. Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2017: 1925-1934. [16] ZHAO Heng-shuang, SHI Jian-ping, QI Xiao-juan, et al. Pyramid scene parsing network[C]//IEEE. Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2017: 2881-2890. [17] CHEN L C, PAPANDREOU G, KOKKINOS I, et al. Deeplab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4): 834-848. doi: 10.1109/TPAMI.2017.2699184 [18] CHEN L C, PAPANDREOU G, SCHROFF F, et al. Rethinking atrous convolution for semantic image segmentation[J]. arXiv, 2017, DOI: 10.48550/arXiv.1706.05587. [19] CHEN L C, ZHU Yu-kun, PAPANDREOU G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//ECCV. Proceedings of the 2018 European Conference on Computer Vision. Munich: ECCV, 2018: 801-818. [20] WANG Jing-dong, SUN Ke, CHENG Tian-heng, et al. Deep high-resolution representation learning for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(10): 3349-3364. doi: 10.1109/TPAMI.2020.2983686 [21] BAI Hai-wei, CHENG Jian, HUANG Xia, et al. HCANet: a hierarchical context aggregation network for semantic segmentation of high-resolution remote sensing images[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 1-5. [22] PANG Shi-yan, SHI Ye-peng, HU Han-chun, et al. PTRSegNet: a patch-to-region bottom-up pyramid framework for the semantic segmentation of large-format remote sensing images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17: 3664-3673. doi: 10.1109/JSTARS.2024.3352578 [23] WANG Xiao-feng, KANG Meng-lei, CHEN Yan, et al. Adaptive local cross-channel vector pooling attention module for semantic segmentation of remote sensing imagery[J]. Remote Sensing, 2023, 15(8): 1980. doi: 10.3390/rs15081980 [24] GUO Yong-jie, WANG Feng, XIANG Yu-ming, et al. Semisupervised semantic segmentation with certainty-aware consistency training for remote sensing imagery[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023, 16: 2900-2914. doi: 10.1109/JSTARS.2023.3255553 [25] SUN Ke, XIAO Bin, LIU Dong, et al. Deep high-resolution representation learning for human pose estimation[C]//IEEE. Proceedings of the 2019 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2019: 5693-5703. [26] HU Jie, SHEN Li, SUN Gang. Squeeze-and-excitation networks[C]//IEEE. Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 7132-7141. [27] LUO Xiang-de, WANG Guo-tai, LIAO Wen-jun, et al. Semi-supervised medical image segmentation via uncertainty rectified pyramid consistency[J]. Medical Image Analysis, 2022, 80: 102517. doi: 10.1016/j.media.2022.102517 [28] CHENG Gong, HAN Jun-wei, LU Xiao-qiang. Remote sensing image scene classification: benchmark and state of the art[J]. Proceedings of the IEEE, 2017, 105(10): 1865-1883. doi: 10.1109/JPROC.2017.2675998 [29] TENG Jing, LI Long-kai, JIANG Ya-jun, et al. A review of clean energy exploitation for railway transportation systems and its enlightenment to China[J]. Sustainability, 2022, 14(17): 10740. doi: 10.3390/su141710740 [30] JIANG Bao-de, AN Xiao-ya, XU Shao-fen, et al. Intelligent image semantic segmentation: a review through deep learning techniques for remote sensing image analysis[J]. Journal of the Indian Society of Remote Sensing, 2023, 51(9): 1865-1878. doi: 10.1007/s12524-022-01496-w -
点击查看大图
图(7) / 表(5)
计量
- 文章访问数: 12
- HTML全文浏览量: 5
- PDF下载量: 3
- 被引次数: 0
