三种图像匹配模型在地图匹配中的适用性

doi:10.13474/j.cnki.11-2246.2024.0819

测绘通报 ›› 2024, Vol. 0 ›› Issue (8): 109-114,121.doi: 10.13474/j.cnki.11-2246.2024.0819

• 学术研究 • 上一篇

三种图像匹配模型在地图匹配中的适用性

侯佳鑫^1,2, 车向红², 刘纪平^1,2, 王鸿雁², 王勇², 徐胜华², 罗安²

1. 兰州交通大学测绘与地理信息学院, 甘肃兰州 730070;
2. 中国测绘科学研究院, 北京 100036

收稿日期:2024-01-29 发布日期:2024-09-03
通讯作者: 刘纪平。E-mail:liujp@casm.ac.cn
作者简介:侯佳鑫(2000—),男,硕士生,主要研究方向为图像处理与应用。E-mail:15244644323@163.com
基金资助:
国家重点研发计划(2022YFC3005705);国家自然科学基金(42071384);地理信息安全监管(A2401)

Applicability of three image matching models in map image matchings

HOU Jiaxin^1,2, CHE Xianghong², LIU Jiping^1,2, WANG Hongyan², WANG Yong², XU Shenghua², LUO An²

1. Faculty of Geomatics, Lanzhou Jiaotong University, Lanzhou 730070, China;
2. Chinese Academy of Surveying and Mapping, Beijing 100036, China

Received:2024-01-29 Published:2024-09-03

摘要/Abstract

摘要： 随着计算机和人工智能等科学技术的快速发展,已出现多种基于深度学习的智能化图像匹配模型,但这些模型大多用于匹配图像纹理特征较为固定的图像(如人物、建筑物和工业器件等),缺乏对图像纹理特征复杂多变的地图图像匹配适用性研究。为此,本文以行政地图、交通地图和专题地图为数据源,基于Superglue、COTR和GlueStick 3种图像匹配深度学习模型,通过融合3类地图和语义分割模型(SAM),从目视检验、匹配精度和匹配效率方面对比分析3种图像匹配模型效果。结果表明:①GlueStick模型在3类地图上边界匹配效果整体最为理想,Superglue次之,而COTR模型匹配效果最差;②基于SAM生成地图分割掩膜图能够减少地图周边干扰要素,进而提升GlueStick和Superglue模型匹配效果,其匹配精度分别提升了47.50%和34.43%;③COTR模型匹配效率最低,GlueStick相比Superglue,在地图原图上匹配效率较低,但在地图分割掩膜图匹配中,两个模型匹配效率相当。本文可为地图内容对比、识别和审查等方面提供重要的智能化技术支撑。

关键词: 地图匹配, 深度学习, Superglue, COTR, GlueStick

Abstract: With the rapid development of science and technology such as computer and artificial intelligence, there are a variety of intelligent image matching models based on deep learning. However, these models are mostly used for matching images with relatively regular texture features (such asportraits, building images,industrial parts, etc), and there is a lack of applicability research in map image matching with complex and diverse texture features. To this end, by taking three types of maps, namely the administrative division map, traffic map and thematic map as data sources, this study compares the map matching performances of the currently popular three image matching models, such as Superglue, COTR and GlueStick, by integrating semantic segmentation model (SAM). The visual matching performance, matching accuracy and matching efficiency results show that:①The boundary matching performance of GlueStick model is the best among the three types of maps, followed by Superglue, while COTR model has the worst matching performance.②Using SAM to extract map segmentation mask images is able to reduce peripheral features of the map, and further improve the matching performance of GlueStick and Superglue models where the accuracy are increased by 47.50% and 34.43%, respectively.③The matching efficiency of the COTR model is the lowest. While the matching efficiency of GlueStick model is lower than Superglue using the original map, their matching efficiency is comparable using the map segmentation masks. This study has important application values for contrast, recognition and review of map content.

Key words: map image matching, deep learning, Superglue, COTR, GlueStick

中图分类号:

P208

侯佳鑫, 车向红, 刘纪平, 王鸿雁, 王勇, 徐胜华, 罗安. 三种图像匹配模型在地图匹配中的适用性[J]. 测绘通报, 2024, 0(8): 109-114,121.

HOU Jiaxin, CHE Xianghong, LIU Jiping, WANG Hongyan, WANG Yong, XU Shenghua, LUO An. Applicability of three image matching models in map image matchings[J]. Bulletin of Surveying and Mapping, 2024, 0(8): 109-114,121.

参考文献

[1] JIANG Xingyu,MA Jiayi,XIAO Guobao,et al. A review of multimodal image matching: methods and applications[J]. Information Fusion,2021,73: 22-71.
[2] 王铮,刘纪平,车向红,等. 基于卷积神经网络的地图相似度匹配方法研究[J]. 测绘科学,2022,47(7): 169-175.
[3] 白又达,刘纪平,黄龙,等. 面向地图图片识别的两种卷积神经网络分析[J]. 测绘科学,2021,46(11): 126-134.
[4] 黄龙. 一种快速挖掘互联网问题地图图片的方法[J]. 测绘与空间地理信息,2017,40(11): 92-93.
[5] 樊凡渊. 对地图的监管不能仅限于纸质版[N]. 中国新闻出版广电报,2019-02-28(3).
[6] 杜凯旋,王亮,王勇,等. 主动学习和卷积神经网络的地图图片识别方法[J]. 测绘科学,2020,45(7): 139-147.
[7] MA Jiayi,JIANG Xingyu,FAN Aoxiang,et al. Image matching from handcrafted to deep features: a survey[J]. International Journal of Computer Vision,2021,129(1): 23-79.
[8] WELLS W M,VIOLA P,ATSUMI H,et al.Multi-modal volume registration by maximization of mutual information[J].Medical Image Analysis,1996,1(1):35-51.
[9] 舒军,李灵,邓明舟. 基于NCC的PCB图像配准算法研究[J]. 湖北工业大学学报,2022,37(2): 48-54.
[10] ROSTEN E,PORTER R,DRUMMOND T. Faster and better: a machine learning approach to corner detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2010,32(1): 105-119.
[11] LOWE D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision,2004,60(2): 91-110.
[12] MATAS J,CHUM O,URBAN M,et al. Robust wide-baseline stereo from maximally stable extremal regions[J].Image and Vision Computing,2004,22(10):761-767.
[13] 陈建新,唐丽玉. 一种SIFT-FREAK图像匹配算法[J]. 测绘与空间地理信息,2023,46(6): 32-35.
[14] 杜玉龙,李建增,张岩,等. 基于MSER与SIFT融合的多通道图像匹配算法研究[J]. 军械工程学院学报,2016,28(1): 52-58.
[15] SARLIN P E,DETONE D,MALISIEWICZ T,et al. Superglue: learning feature matching with graph neural networks[J].Computer Science,2019:208291327.
[16] PAUTRAT R,SUÁREZ I,YU Yifan,et al. GlueStick: robust image matching by sticking points and lines together[C]//Proceedings of 2023 IEEE/CVF International Conference on Computer Vision (ICCV). Paris: IEEE,2023: 9706-9716.
[17] CAO Xiaohuan,YANG Jianhuan,WANG Li,et al. Deep learning based inter-modality image registration supervised by intra-modality similarity[M]//SHI Yinghuan,SUK H I,LIU Mingxia,etc. Machine Learning in Medical Imaging. Cham: Springer International Publishing,2018: 55-63.
[18] MAHAPATRA D,ANTONY B,SEDAI Suman,et al. Deformable medical image registration using generative adversarial networks[C]//Proceedings of 2018 IEEE International Symposium on Biomedical Imaging. Washington,DC: IEEE,2018: 1449-1453.
[19] JIANG Wei,TRULLS E,HOSANG J,et al. COTR: correspondence transformer for matching across images[C]//Proceedings of 2021 IEEE/CVF International Conference on Computer Vision.Montreal,QC,Canada: IEEE,2021: 6207-6217.
[20] KIRILLOV A,MINTUN E,RAVI N,et al. Segment anything[J].Computer Science,2023:257952310.
[21] HUANG Y,YANG X,LIU L,et al. Segment anything model for medical images?[J]. Med Image Anal,2024,92: 103061.
[22] JI G P,FAN D P,XU P,et al. SAM struggles in concealed scenes--empirical study on “segment anything”[J]. Computer Science,2023:258079139.
[23] TANG Lü,XIAO Haoke,LI Bo. Can SAM segment anything? when SAM meets camouflaged object detection[J]. Computer Science,2023:258048579.

三种图像匹配模型在地图匹配中的适用性

Applicability of three image matching models in map image matchings

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	石一剑, 谭海, 钟旭辉. 改进YOLOv5的国产光学影像辐射异常检测方法[J]. 测绘通报, 2024, 0(8): 60-65,72.
[2]	尤相骏, 赵霞, 龙四春, 王嘉伟, 郑颖, 邝利军. 地铁隧道病害检测深度学习模型优化及应用[J]. 测绘通报, 2024, 0(8): 96-101.
[3]	马下平, 王风凯, 赵庆志, 高余婷. 深度学习的大高差高海拔地区高程拟合方法[J]. 测绘通报, 2024, 0(8): 102-108.
[4]	杨晓峰, 魏浩翰, 张强, 向云飞. 融合孤立森林和深度学习的GNSS-IR土壤湿度反演[J]. 测绘通报, 2024, 0(7): 88-94.
[5]	田亚林, 连增增, 王鹏辉, 王孟奇, 陆力. 基于KF-LSTM的UWB室内定位算法[J]. 测绘通报, 2024, 0(7): 95-99,151.
[6]	刘兵, 刘宇, 金凤学, 邹一波, 葛艳, 赵林林. 针对浒苔目标检测的全局背景强化的位置蒸馏方法[J]. 测绘通报, 2024, 0(6): 19-23.
[7]	高墨通, 杨维芳, 刘祖昱, 曹小双, 张瑞琪, 侯宇豪. 结合卷积神经网络和注意力机制的LSTM采空区地表沉降预测方法[J]. 测绘通报, 2024, 0(6): 53-58,170.
[8]	钟青岑, 吴晨昊, 向隆刚, 姚鹏. 顾及路网约束的深度地图匹配方法[J]. 测绘通报, 2024, 0(6): 96-102,133.
[9]	尹泽中, 李功权. 三维场叠加视频流的电子围栏越界检测方法[J]. 测绘通报, 2024, 0(6): 103-108.
[10]	王丽霞, 喜文飞, 史正涛, 赵子龙, 钱堂慧, 赵磊, 马奕捷. 深度学习滑坡识别算法中样本不平衡问题的研究[J]. 测绘通报, 2024, 0(5): 12-18.
[11]	姚楚羡, 蔡皓楠, 张远波, 唐可懿, 詹璐, 周宝定. 基于轻量化车载设备的道路病害检测方法[J]. 测绘通报, 2024, 0(5): 147-150.
[12]	张楠, 张蕴灵. 基于深度学习模型的复杂困难路网提取方法[J]. 测绘通报, 2024, 0(5): 155-159.
[13]	赵明姝. 一种更新高精地图置信度的方法[J]. 测绘通报, 2024, 0(5): 166-171,177.
[14]	林泊锟, 丁勇, 李登华. 一种基于深度学习与图像局部特征提取的边坡异常监测技术[J]. 测绘通报, 2024, 0(4): 23-28.
[15]	庞敏. 基于深度学习的国产卫片图斑提取[J]. 测绘通报, 2024, 0(4): 124-128,134.