A Stereo Selecting Method of Multi-view Matching Models Guided Based on Feature Points
-
摘要: 从冗余数据中选择一个或者多个最为显著的立体像对，在最少“伪信息”的影响下，获取最佳影像匹配效果，降低其它质量较差影像的负面平均效应，是提高多视影像匹配性能的关键。基于准确匹配的特征点，通过匹配测度的鲁棒性分析，提出一种多视影像的匹配质量分析方法；在此基础上，提出了一种基于特征点引导的多视影像择优匹配方法及基本思想、计算基础和择优匹配步骤。利用ADS40多度重叠影像数据进行了择优匹配实验。结果表明，该方法能够有效选取匹配质量较优的影像，获取更加准确的多视匹配结果，在一定程度上，比传统的多视匹配方法更加有效。Abstract: Selecting one or more robust matching stereo pairs from redundant overlapping images to reduce the negative influence of incorrect or confusing image information to obtain the most desirable matching results improves multi-view matching ability and quality in multi-view matching techniques. In this paper, a matching quality analysis method for multi-view images is proposed that measures matching robustness based on correctly matched SIFT feature points. Furthermore, based on the method, a feature point guided multi-view image stereo selection matching method is detailed including the basic principles, algorithm, and matching process. Experiments were done on ADS40 multi-view imagery. The results show that the proposed method automatically and efficiently selects images of high matching quality from redundant overlapping images to obtain more correct multi-view matching results. This method is, to some extent, more effective than traditional multi-view matching methods.
-
Keywords:
- feature point /
- multi-view image /
- matching quality /
- matching measure /
- stereo selecting matching
-
-
![]()
图 1 待匹配点的邻近特征点搜索
Figure 1. Neighboring Feature Points Searching Method for Matching Point
![]()
图 2 SIFT特征点匹配结果图
(圆形框内为兴趣区域，十字丝为待匹配的像点)
Figure 2. SIFT Matching Results(Circle Area Was Interest Area, "+"Refers to a Point to be Matched)
![]()
图 6 仅使用质量较优的搜索影像时待匹配点P0择优匹配测度曲线
Figure 6. Stereo Selecting and Matching Measure Curve of Point P0 when Images of Great Matching Quality were Used Only
![]()
图 8 待匹配点P0的GC3多视匹配模型像方搜索范围
(N1为基准影像，其它影像为搜索影像；“+”为初始匹配点；线段为像方搜索范围；“×”为同名匹配点)
Figure 8. Image Searching Space of Matching Point P0 by GC3 Multi-view Matching Model
(N1 was Reference Image, the others were Searching Images; "+"was Initial Matching Result; Lines Referred to Searching Space; "×"was successfully matched points for P0)
![]()
图 7 使用更多的搜索影像时待匹配点P0择优匹配测度曲线
Figure 7. Stereo Selecting and Matching Measure Curve of Point P0 when more Images of Great Matching Quality were Used
表 1 多视影像的匹配质量分析
Table 1 Matching Quality Analysis of Multi-view Images
匹配质量类型 匹配测度分析一 匹配测度分析二 质量标识 匹配质量 (1) ρi1＞0.90 ρi1/ρi2＞1.2 1 较优 ρi1＞0.80 ρi1/ρi2＞1.4 1 较优 ρi1＞0.65 ρi1/ρi2＞1.6 1 较优 (2) ρi1＞0.90 ρi1/ρi2＜1.2 0 一般 ρi1＞0.80 ρi1/ρi2＜1.4 0 一般 ρi1＞0.65 ρi1/ρi2＜1.6 0 一般 0.50＜ρi1＜0.65 ρi1/ρi2＞1.4 0 一般 (3) 0.50＜ρi1＜0.65 ρi1/ρi2＜1.4 -1 较差 ρi1＜0.50 / -1 较差 注：沿匹配方向线的相关测度曲线ρi(ρi1最大相关系数，ρi2次局部最大相关系数，单位为1) 
下载: 导出CSV
表 2 ADS40实验数据参数
Table 2 Parameters of ADS40 Experiment Image
数据 预处理级 焦距 地面采样间隔 相对航高 影像数量 数据航带 ADS40 L1级 62.5 mm 0.21 m 2 000 m 6张 2条
下载: 导出CSV
表 3 搜索影像的匹配质量分析与标识
Table 3 Matching Quality Analysis and Index of Searching Images
匹配测度及分析 N1-B1 N1-F1 N1-B2 N1-N2 N1-F2 特征点P1 ρ1 0.908 3 0.780 1 0.768 3 0.677 1 0.767 0 ρ2 0.714 2 0.738 7 0.441 2 0.530 6 0.576 9 ρ1/ρ2 1.271 8 1.056 0 1.741 3 1.276 1 1.329 5 质量标识 1 0 1 0 0 特征点P2 ρ1 0.957 2 0.888 7 0.943 5 0.872 5 0.969 2 ρ2 -0.243 2 0.569 1 0.150 0 0.371 9 0.446 7 ρ1/ρ2 3.935 9 1.561 6 6.290 0 2.346 1 2.169 7 质量标识 1 1 1 1 1 特征点P3 ρ1 0.762 8 0.655 7 0.398 9 0.297 7 0.329 3 ρ2 0.553 9 0.266 8 0.325 9 0.220 8 0.321 0 ρ1/ρ2 1.377 1 2.457 6 1.224 0 1.348 3 1.025 9 质量标识 0 1 -1 -1 -1
下载: 导出CSV
表 4 搜索影像的匹配质量标识累加结果
Table 4 Cumulative Matching Quality of Searching Images
N1-B1 N1-F1 N1-B2 N1-N2 N1-F2 质量标识累加 2 2 1 0 0
下载: 导出CSV
-
[1] 张永生, 范大昭, 纪松.用于ADS40传感器的多视觉立体匹配算法模型[J].测绘科学技术学报, 2007, 24(2):83-86 Zhang Yongsheng, Fan Dazhao, Ji Song. Multi-view Matching Algorithm Model for ADS40 Sensor[J].Journal of Zhengzhou Institute of Surveying and Mapping, 2007, 24(2):83-86
[2] 姚国标, 邓喀中, 杨化超等.基于几何约束的倾斜立体影像匹配[J].中国矿业大学学报, 2013, 42(4):676-682 http://industry.wanfangdata.com.cn/dl/Detail/Periodical?id=Periodical_zgkydxxb201304024 Yao Guoji, Deng Kazhong, Yang Huachao, et al. Matching Methods for Oblique Stereo Images on the Basis of Geometry Constraint[J]. Journal of China University of Mining & Technology, 2013, 42(4):676-682 http://industry.wanfangdata.com.cn/dl/Detail/Periodical?id=Periodical_zgkydxxb201304024
[3] 柯涛, 张祖勋, 张剑清.旋转多基线数字近景摄影测量[J].武汉大学学报·信息科学版, 2009, 34(1):44-47 http://ch.whu.edu.cn/CN/abstract/abstract1135.shtml Ke Tao, Zhang Zuxun, Zhang Jianqing. Panning and Multi-baseline Digital Close-range Photogrammetry[J].Geomatics and Information Science of Wuhan University, 2009, 34(1):44-47 http://ch.whu.edu.cn/CN/abstract/abstract1135.shtml
[4] 纪松. 多视匹配策略与优化方法研究[D]. 郑州: 信息工程大学, 2012 http://cdmd.cnki.com.cn/Article/CDMD-90005-1013161264.htm Ji Song. Study on The Strategy and Improvement Method of Multi-view Matching Technology[D]. Zhengzhou: Information Engineering University, 2012 http://cdmd.cnki.com.cn/Article/CDMD-90005-1013161264.htm
[5] 刘经南, 方媛, 郭迟等.位置大数据的分析处理研究进展[J].武汉大学学报·信息科学版, 2014, 39(4):379-385 http://ch.whu.edu.cn/CN/abstract/abstract2947.shtml Liu Jingnan, Fang Yuan, Guo Chi, et al. Reaserch Progress in Location Big Data Analysis and Processing[J].Geomatics and Information Science of Wuhan University, 2014, 39(4):379-385 http://ch.whu.edu.cn/CN/abstract/abstract2947.shtml
[6] Zhang B C. Towards a Higher Level of Automation in Softcopy Photogrammetry: NGATE and LiDAR Processing in SOCET SET®[C]. GeoCue Corporation 2nd Annual Technical Exchange Conference, Tennessee, USA, 2006
[7] Rumpler M, Irschara A, Bischof H. Multi-view Stereo: Redundancy Benefits for 3D Reconstruction[C]. Proceedings of the 35th Workshop of the Austrian Association for Pattern Recognition, Graz, Austria, 2011 http://www.researchgate.net/publication/228931092_Multi-View_Stereo_Redundancy_Benefits_for_3D_Reconstruction
[8] Alsadik B, Gerke M, Vosselman G. Optimal Camera Network Design for 3D modeling of Cultural Heritage Objects[J]. Journal of Cultural Heritage, 2012, 14(6):515-526 https://www.researchgate.net/profile/Bashar_Alsadik/publication/261158512_Automatic_image_network_design_leading_to_optimal_image-based_3D_models/links/0f31753359712cc446000000.pdf?inViewer=true&disableCoverPage=true&origin=publication_detail
[9] Ahmadabadian A H, Robson S, Boehm J, et al. Image Selection in Photogrammetric Multi-View Stereo Methods for Metric and Complete 3D Reconstruction[C]. Videometrics, Range Imaging and Applications Conference, Munich, Germany, 2013 doi: 10.1117/12.2020472
[10] Zhang L. Automatic Digital Surface Model (DSM) Generation from Linear Array Images[D]. Zurich: Swiss Federal Institute of Technology, 2005 http://www.researchgate.net/publication/36180044_Automatic_Digital_Surface_Model_DSM_Generation_from_Linear_Array_Images
[11] 张力, 张继贤.基于多基线影像匹配的高分辨率遥感影像DEM的自动生成[J].武汉大学学报·信息科学版, 2008, 33(9):943-946 http://ch.whu.edu.cn/CN/abstract/abstract1689.shtml Zhang Li, Zhang Jixian. Automatic DEM Generation from High-Resolution Satellite Imagery Based on Multiple-Baseline Image Matching[J].Geomatics and Information Science of Wuhan University, 2008, 33(9):943-946 http://ch.whu.edu.cn/CN/abstract/abstract1689.shtml
[12] 李晓明, 郑链, 胡占义.基于SIFT特征的遥感影像自动配准[J].遥感学报, 2006, 6(10):885-892 http://d.wanfangdata.com.cn/Periodical/ygxb200606009 Li Xiaoming, Zheng Lian, Hu Zhanyi. SIFT Based Automatic Registration of Remotely-sensed Imagery[J].Journal of Remote Sensing, 2006, 6(10):885-892 http://d.wanfangdata.com.cn/Periodical/ygxb200606009
[13] 杨化超, 姚国标, 王永波.基于SIFT的宽基线立体影像密集匹配[J].测绘学报, 2011, 40(5):537-543 http://www.cqvip.com/QK/90069X/201105/39558505.html Yang Huachao, Yao Guoji, Wang Yongbo. Dense Matching for Wide Base-line Stereo Images Based on SIFT[J]. Acta Geodaetica et Cartographica Sinica, 2011, 40(5):537-543 http://www.cqvip.com/QK/90069X/201105/39558505.html
[14] 纪松, 范大昭, 张永生等.多视匹配MVLL算法及其在ADS40线阵影像中的运用[J].武汉大学学报·信息科学版, 2009, 34(1):28-31 http://ch.whu.edu.cn/CN/abstract/abstract1127.shtml Ji Song, Fan Dazhao, Zhang Yongsheng, et al. MVLL Multi-view Matching Algorithm and its Application on ADS40 Liner Array Imagery[J]. Geomatics and Information Science of Wuhan University, 2009, 34(1):28-31 http://ch.whu.edu.cn/CN/abstract/abstract1127.shtml
-
期刊类型引用(3)
1. 周永章，陈川，张旗，王功文，肖凡，沈文杰，卞静，王亚，杨威，焦守涛，刘艳鹏，韩枫. 地质大数据分析的若干工具与应用. 大地构造与成矿学. 2020(02): 173-182 . 
百度学术
2. 王叶晨梓，杜震洪，张丰，刘仁义. 面向分片地图的多分辨率格点数据统一存取方法. 浙江大学学报(理学版). 2017(05): 584-590 . 百度学术
3. 朱建章，石强，陈凤娥，史晓丹，董泽民，秦前清. 遥感大数据研究现状与发展趋势. 中国图象图形学报. 2016(11): 1425-1439 . 百度学术
其他类型引用(6)
计量
- 文章访问数: 2293
- HTML全文浏览量: 191
- PDF下载量: 351
- 被引次数: 9
