基于区域候选网络的矢量线要素自动化简方法

doi:10.11947/j.AGCS.2023.20220363

摘要/Abstract

摘要： 针对现有矢量线要素化简算法智能化程度不高的问题,本文提出了一种基于区域候选网络的矢量线要素自动化简方法。首先利用深度可分离卷积神经网络实现矢量线要素栅格化后的卷积特征提取;然后结合线要素的坐标信息改进区域候选网络中候选框的生成方式,以实现候选框与可能的弯曲组合的对应;最后在统一线要素弯曲特征图大小的基础上,根据候选框对应的卷积特征进行二分类判断,完成弯曲单元的自动识别,并通过删除弯曲单元实现线化简。本文在海岸线数据集上进行了模型的训练和测试,并通过不同栅格化参数、不同骨干网络、跨尺度化简,以及不同类型线要素化简等对比试验验证了模型的有效性。试验结果表明,本文方法能够自动学习已有线化简案例中的化简知识,并充分利用线要素的矢量特征及栅格特征,自动完成线要素弯曲特征的准确定位,最终实现可端到端训练的矢量线要素自动化简。

关键词: 线化简, 区域候选网络, 地图综合, 深度学习

Abstract: To address the problem of existing vector polyline simplification algorithms lacking intelligence, an automatic vector polyline simplification algorithm based on regional proposal network is proposed. Firstly, a depth-separable convolutional neural network is used to realize the convolutional feature extraction of raster polylines. Then, the generation method of region proposals in the region proposal network is improved by combining the coordinate information of the polylines, which is in order to establish the correspondence between the proposal region and the possible bending combinations. Finally, after unifying the size of the bending feature map, the automatic detection of bending units is completed by binary classification according to the convolutional features corresponding to the region proposals, and polyline simplification is achieved by deleting the bending units. In this paper, the model is trained and tested on the coastline dataset, and the effectiveness of the model is verified by comparison experiments with different rasterization parameters, different backbone networks, cross-scale simplification, and different types of line simplification. Experimental results show that this algorithm can intelligently learn the simplification knowledge from the existing polyline simplification cases, and make full use of the vector and raster features of polylines to identify and locate the bending units, and finally complete the automatic simplification of vector polylines while the network can be trained end-to-end.

Key words: line simplification, region proposal network, map generalization, deep learning

中图分类号:

P283

江宝得, 许少芬, 巫勇, 王淼. 基于区域候选网络的矢量线要素自动化简方法[J]. 测绘学报, 2023, 52(12): 2209-2222.

JIANG Baode, XU Shaofen, WU Yong, WANG Miao. Automatic vector polyline simplification based on region proposal network[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(12): 2209-2222.

参考文献

[1] COURTIAL A, EL AYEDI A, TOUYA G, et al. Exploring the potential of deep learning segmentation for mountain roads generalisation[J].ISPRS International Journal of Geo-Information, 2020, 9(5): 338.
[2] YU Wenhao, CHEN Yujie. Data-driven polyline simplification using a stacked autoencoder-based deep neural network[J]. Transactions in GIS, 2022, 26(5): 2302-2325.
[3] SONG Jia, MIAO Ru. A novel evaluation approach for line simplification algorithms towards vector map visualization[J]. ISPRS International Journal of Geo-Information, 2016, 5(12): 223.
[4] ZHANG Lihua, TANG Lulu, JIA Shuaidong, et al. A collaborative simplification method for multiple coastlines based on the hierarchical triangulation network partition[J]. Journal of Geodesy and Geoinformation Science, 2020, 3(2): 93-104.
[5] DOUGLAS D H, PEUCKER T K. Algorithms for the reduction of the number of points required to represent a digitized line or its caricature[J]. Cartographica: the International Journal for Geographic Information and Geovisualization, 1973, 10(2): 112-122.
[6] LI Zhilin, OPENSHAW S. Algorithms for automated line generalization based on a natural principle of objective generalization[J]. International Journal of Geographical Information Systems, 1992, 6(5): 373-389.
[7] VISVALINGAM M, WHYATT J D. Line generalisation by repeated elimination of points[J]. The Cartographic Journal, 1993, 30(1): 46-51.
[8] TEH C H, CHIN R T. On the detection of dominant points on digital curves[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1989, 11(8): 859-872.
[9] WANG Zeshen, MVLLER J C. Line generalization based on analysis of shape characteristics[J]. Cartography and Geographic Information Systems, 1998, 25(1): 3-15.
[10] 艾廷华. 深度学习赋能地图制图的若干思考[J]. 测绘学报, 2021, 50(9): 1170-1182. DOI: 10.11947/j.AGCS.2021.20210091. AI Tinghua. Some thoughts on deep learning enabling cartography[J]. Acta Geodaetica et Cartographica Sinica, 2021, 50(9): 1170-1182. DOI: 10.11947/j.AGCS.2021.20210091.
[11] KENT A. Trust me, I'm a cartographer: post-truth and the problem of acritical cartography[J]. The Cartographic Journal, 2017, 54(3): 193-195.
[12] 杜佳威, 武芳, 朱丽, 等. 图形、图像融合利用的集成学习智能化简方法及其在岛屿岸线化简中的应用[J]. 测绘学报, 2022, 51(3): 373-387. DU Jiawei, WU Fang, ZHU Li, et al. An ensemble learning simplification approach based on multiple machine-learning algorithms with the fusion using of raster and vector data and a use case of coastline simplification[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(3): 373-387.
[13] 何海威, 钱海忠, 段佩祥, 等. 线要素化简及参数自动设置的案例推理方法[J]. 武汉大学学报(信息科学版), 2020, 45(3): 344-352. HE Haiwei, QIAN Haizhong, DUAN Peixiang, et al. Automatic line simplification algorithm selecting and parameter setting based on case-based reasoning[J]. Geomatics and Information Science of Wuhan University, 2020, 45(3): 344-352.
[14] 武芳, 杜佳威, 钱海忠, 等. 地图综合智能化研究的发展与思考[J]. 武汉大学学报(信息科学版), 2022, 47(10): 1675-1687. WU Fang, DU Jiawei, QIAN Haizhong, et al. Overview of research progress and reflections in intelligent map generalization[J]. Geomatics and Information Science of Wuhan University, 2022, 47(10): 1675-1687.
[15] KARSZNIA I, SIELICKA K. When traditional selection fails: how to improve settlement selection for small-scale maps using machine learning[J]. ISPRS International Journal of Geo-Information, 2020, 9(4): 230.
[16] KARSZNIA I, WEIBEL R. Improving settlement selection for small-scale maps using data enrichment and machine learning[J]. Cartography and Geographic Information Science, 2018, 45(2): 111-127.
[17] SHEN Yilang, AI Tinghua, HE Yakun. A new approach to line simplification based on image processing: a case study of water area boundaries[J]. ISPRS International Journal of Geo-Information, 2018, 7(2): 41.
[18] PARK W, YU K. Hybrid line simplification for cartographic generalization[J]. Pattern Recognition Letters, 2011, 32(9): 1267-1273.
[19] 段佩祥, 钱海忠, 何海威, 等. 基于支持向量机的线化简方法[J]. 武汉大学学报(信息科学版), 2020, 45(5): 744-752, 783. DUAN Peixiang, QIAN Haizhong, HE Haiwei, et al. A line simplification method basedon support vector machine[J]. Geomatics and Information Science of Wuhan University, 2020, 45(5): 744-752, 783.
[20] SESTER M, FENG Y, THIEMANN F. Building generalization using deep learning[J]. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2018, XLII-4: 565-572.
[21] TOUYA G, ZHANG Xiang, LOKHAT I. Is deep learning the new agent for map generalization?[J]. International Journal of Cartography, 2019, 5(2/3): 142-157.
[22] YANG Min, YUAN Tuo, YAN Xiongfeng, et al. A hybrid approach to building simplification with an evaluator from a backpropagation neural network[J]. International Journal of Geographical Information Science, 2022, 36(2): 280-309.
[23] RONNEBERGER O, FISCHER P, BROX T. U-net: convolutional networks for biomedical image segmentation[C]//Proceedings of 2015 International Conference on Medical Image Computing and Computer-Assisted Intervention. Munich: IEEE, 2015: 234-241.
[24] DU Jiawei, WU Fang, XING Ruixing, et al. Segmentation and sampling method for complex polyline generalization based on a generative adversarial network[J]. Geocarto International, 2022, 37(14): 4158-4180.
[25] ISOLA P, ZHU Junyan, ZHOU Tinghui, et al. Image-to-image translation with conditional adversarial networks[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu:IEEE, 2017: 5967-5976.
[26] 江宝得, 吴亮, 谢忠. 顾及海岸线地理弯曲特征约束的可控分形插值方法[J]. 武汉大学学报(信息科学版), 2019, 44(3): 451-458. JIANG Baode, WU Liang, XIE Zhong. A controlled fractal interpolation method for coastline considering bending characteristic constraints[J]. Geomatics and Information Science of Wuhan University, 2019, 44(3): 451-458.
[27] SANDLER M, HOWARD A, ZHU Menglong, et al. MobileNetV2: inverted residuals and linear bottlenecks[EB/OL].[2022-05-29]. https://arxiv.org/pdf/1801.04381.pdf.
[28] REN Shaoqing, HE Kaiming, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[29] GIRSHICK R. Fast R-CNN[C]//Proceedings of 2015 IEEE International Conference on Computer Vision(ICCV 2015). Santiago: IEEE, 2015: 1440-1448.
[30] CHEN Guanzhou, ZHANG Xiaodong, WANG Qing, et al. Symmetrical dense-shortcut deep fully convolutional networks for semantic segmentation of very-high-resolution remote sensing images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2018, 11(5): 1633-1644.
[31] WESSEL P, SMITH W H F. A global, self-consistent, hierarchical, high-resolution shoreline database[J]. Journal of Geophysical Research: Solid Earth, 1996, 101(B4): 8741-8743.
[32] CHANG Zihan, ZHANG Yang, CHEN Wenbo. Effective adam-optimized LSTM neural network for electricity price forecasting[C]//Proceedings of 2018 International Conference on Software Engineering and Service Science. Beijing: IEEE, 2018: 245-248.
[33] MCMASTER R B. A statistical analysis of mathematical measures for linear simplification[J]. The American Cartographer, 1986, 13(2): 103-116.
[34] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL].[2022-05-29]. https://arxiv.org/abs/1409.1556.pdf.
[35] LIN T Y, DOLLAR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 936-944.
[36] 江宝得, 黄威, 许少芬, 等. 融合分散自适应注意力机制的多尺度遥感影像建筑物实例细化提取[J]. 测绘学报,2023,52(9):1504-1514. DOI: 10.11947/j.AGCS.2023.20220322. JIANG Baode, HANG Wei, XU Shaofen, et al. Multi-scale building instance refinement extraction from remote sensing images by fusing with decentralized adaptive attention mechanism[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(9): 1504-1514. DOI: 10.11947/j.AGCS.2023.20220322.