基于卫星图像的虚拟数字化城市建模
详细信息    本馆镜像全文|  推荐本文 |  |   获取CNKI官网全文
摘要
随着当前遥感技术、计算机技术等不断发展,遥感图像已经成为获取地理信息的重要数据来源。与传统的方法相比较,从遥感图像中获取信息有时效性、周期性、经济性等优势。利用遥感卫星图像获取道路、房屋等地物信息已成为一个备受世界各国关注的问题。道路及房屋是数字城市化的重要组成部分,随着科技及计算机技术的不断发展,快速自动获取质量高、规模大的道路网及房屋信息已成为可能。通过遥感卫星图像提取地物信息已成为计算机视觉、计算机图像处理及遥感等领域的研究热点。
     本论文研究内容涉及图像处理,模式识别和虚拟现实多课题。文中对数字图像处理的一些基本方法进行了讨论,包括图像灰度化、图像二值化、图像增强、图像平滑处理、图像边缘检测、分割算法、直方图统计、腐蚀与膨胀、斑块统计等。
     具体研究内容:
     1.在预处理阶段,探讨了图像质量改善的方法。着重对条带噪声去除展开研究,用插值法、傅里叶变换法去除条带噪声。
     2.探讨了现有边缘检测算子,在获得道路边缘特征的基础上,再用曲线拟合方法对道路边缘进行拟合,实现道路提取。该方法对于乡村道路提取效果较好,但对错综复杂的道路网有待进一步研究。
     3.在图像分割获得道路轮廓的基础上,提出了用方向模板来检测道路。该方法需人工给出道路起始点和终止点,算出两点连线构成的方向角,根据方向角选择对应的方向模板,并从起始点逐步检测到终止点,实现道路提取。该方法需多次给出道路起始点和终止点,如何提高该算法的自动化程度有待进一步研究。
     4.研究了用图像卷积运算获得线特征加强系数,该系数能加强道路线特征,同时也能弱化背景灰度恒定区域。去除短线段和块状噪声,人工干预进行道路修剪,最后用数学形态学细化处理得到道路骨架,实现道路提取。因为道路一般与村庄、城镇等居民地或人工设施相连接,所以该方法需要人工截断道路和居民地的连接。如果能够智能地判断出非道路区域,并实现自动截除,那么道路提取的自动化程度将会得到很大的提高。
     5.在图像分割获得基本道路网轮廓的基础上,提出了用Hough直线,再进行道路判断、道路修剪、道路连接,并形成道路网,实现了城市直线道路提取。该方法对于城市直线道路提取效果较好,但算法鲁棒性有待加强。
     6.提出了基于概率的建筑物轮廓识别算法。在这种算法中,提出了剪切-融合算法用于轮廓的提取与合并,并建立了一个概率模型用以衡量某个轮廓区域是否为建筑物的可能性。这种算法的优点是能够通过概率模型把建筑物的各个特征结合起来,并通过学习由先验数据来决定模型参数,同时,模型也具有很好的扩展性,新的特征很容易加入进来。相对于其他建筑物提取算法,本算法具有更好的鲁棒性和可扩展性。
     7.提出了基于轮廓提取的建筑物变化检测算法。此算法的主要思想是首先找出不同时相图像中的建筑物轮廓,然后利用我们提出的概率模型比较对应的建筑物轮廓并给出它们变化的可能性,最后根据建筑物相应的变化概率值,判断它们是否发生了变化。通过实例可以看到这种算法对于一般纹理复杂度的建筑物的变化具有较好的检测能力,并且算法本身对图像的配准精度要求不高,具有较强的鲁棒性。图[15]参[65]
With the current remote sensing technology and computer technology development, remote sensing image has become the important data sources of geographic information. Compared with the traditional method, remote sensing image for information has the advantages of timeliness periodic, economic and other advantages. Using remote sensing satellite image to access roads and houses Geophysics information has become a concern of the countries around the world. Roads and buildings are important parts of digital urbanization, along with the development of the science and computer technology, quick automatic obtaining high quality and large network information of roads and housing has become possible. Through remote sensing satellite image extracting features information has become the research focus in the field of computer vision, computer image processing, and remote sensing and so on.
     This thesis research content involves image processing, pattern recognition and virtual reality. Some basic methods of digital image processing are discussed in the article, including image gray-scale melt, the binary image, image enhancement, the image smooth processing, the image edge detection, segmentation algorithm, histogram statistics, corrosion and expansive, patch statistics, etc.
     Specific research content:
     1. In the pretreatment stage, the methods to improve image quality are discussed, focusing on the strip noise removal, launched research with interpolation method, flourier transform method removing noise bands.
     2. On the basis of achieving road edge character, the existing edge detection operators and with curve fitting method of road edges fitting to realize road extraction are addressed in the article. This method for rural road extraction effect is better, but for the complex network of roads needs further research.
     3. On the basis of getting road contour through Image segmentation, using the template to detect the direction of the road is proposed. This method calculate the two attachment direction Angle based on the road start and end points, according to the direction of direction Angle choose corresponding template, test the starting point to the end point gradually, realize road extraction.This method requires multiple road start and end points, how to improve the degree of automation algorithm needs further study.
     4. Using image convolution operation to gain line features is researched, the coefficient can strengthen the way line features, also can weaken background gray constant area. Remove short line segment and massive noise, manual intervention for road clip, finally use mathematical morphology refining processing for getting road skeleton, and realize road extraction. Because the road usually connected with villages and towns, so the method needs to manually cut off the connection of roads and residents. If you can intelligently determine the off-road areas, and automatically cut away, then the automation of road extraction will be greatly improved.
     5. On the basis of basic outline of the road network by Image segmentation, proposed by Hough Line, the road again, road cutting, road connections, and the formation of the road network is determined to achieve the city straight road extraction. The line for urban road extraction method is better, but the algorithm robustness needs to be strengthened.
     6. The building contour detection algorithm based on probability is proposed.In this algorithm, cut-fusion algorithm for contour extraction and merge is proposed, the establishment of a probability model to measure the contour area is a possibility of the building. This algorithm has the advantage of combining the various features of the building through probability models, and by learning from the prior data to determine the model parameters, meanwhile, the model also has good scalability, the new feature is easy to join. Compared to other building extraction algorithms, this algorithm has better robustness and scalability.
     7. Building change detection algorithm based on contour extraction is proposed. The main idea of this algorithm is to first find out the image at different phases of the building outline, and then we propose a probabilistic model comparing the corresponding changes in the building outline and give them the possibility of changes in the last according to the probability of the corresponding value of the building to determine whether they have changed. An example can be seen that the algorithm for general texture changes in the building has better detection capabilities, and the algorithm itself is less precision image registration, has strong robustness. Figure [15] reference [65]
引文
[1]赵春江.C#数字图像处理算法典型实例.第1版.北京:人民邮电出版社,2009:236-342
    [2]Gonzalez RC, Woods RE数字图像处理.第2版.北京:电子工业出版社,2003:234-256
    [3]林开颜等.彩色图像分割方法综述[J].中国图象图形学报,2005-10-16(1):159~159
    [4]高守传,姚领田Visual c++数字图像处理与工程应用篇.第2版.北京:中国铁道出版社,2006:135-245
    [5]章毓晋.图像工程(上册)-图像处理和分析[M].第1版.北京:清华大学出版社,1999:135-234
    [6]勒中鑫.数字图像信息处理.第1版.北京:国防工业出版社,2003:235-345
    [7]麦特尔.现代数字图像处理[M].第1版.北京:电子工业出版社,2006:156-234
    [8]Castleman K R. Digital Image Processing[M]北京:清华大学出版社,1998:23-56
    [9]朱长青.基于形态分割的高分辨率遥感影像道路提取.测绘学报,2004(4):348-351
    [10]李利伟等.基于数学形态学的高分辨率遥感影像道路提取.武汉:武汉大学,2005:9-10
    [11]王天鹏.遥感影像中道路的半自动提取研究.郑州:解放军信息工程大学,2004:1-2
    [12]吕健刚.遥感影像道路提取方法研究.桂林:桂林理工大学,2009:9-15
    [13]刘芳.遥感影像中道路自动提取方法研究.北京:北京建筑工程学院.2007:1-6
    [14]魏敏,李朝峰.基于模糊连接度的卫星图像道路提取新方法.计算机工程与应用,2006(]3):230-232
    [15]闫冬梅.基于特征融合的遥感影像典型线状目标提取技术研究.北京:中国科学院研究生院,2003:30-78
    [16]陈卫荣等.基于特征融合的高分辨率SAR图像道路提取.遥感技术与应用学报,2004:30-45
    [17]曹广真,金亚秋.基于水平集方法的多源遥感数据融合及城区道路提取.电子与信息学报,2007:1465-1466
    [18]王莉莉等.基于D-S证据理论的城市航拍道路提取方法.软件学报,2005(2):1535~1538
    [19]章孝灿.遥感数字图像处理.第1版.杭州:浙江大学出版社,1997:125-235
    [20]廖明生等.基于典型相关分析的多元变化检测.遥感学报,2000:197-201
    [21]边肇祺,张学工等.模式识别.第1版.北京:清华大学出版社,1999:23
    [22]汤国安等.遥感数字图像处理.第1版.北京:科学出版社,2004:104
    [23]谢谦礼等.一种基于高分辨遥感影像的道路提取方法.北京:北京大学遥感所,2006:188-189
    [24]崔屹.图像处理与分析——数学形态学方法及应用.第1版.北京:科学出版社,2000:47-47
    [25]孙颖,张志佳.基于频域滤波的自适应条带噪声去除算法.沈阳:沈阳大学,2006:2-3
    [26]陈劲松等.中分辨率遥感图像条带噪声的去除.遥感学报,2004(3):227-229
    [27]瞿洋,杨利平.Hough变换OCR图像倾斜矫正方法.中国图象图形学报, 2001-2-16(2):177-178
    [28]朱长青等.基于结点扩展的高分辨率遥感影像道路提取.华中科技大学学报(自然科学版),2009(1):27-29
    [29]肖志强,鲍光淑.基于GA的SAR图像中主干道路提取.中国图象图形学报,2004-1(1):93-97
    [30]汤国安等.遥感数字图像处理.第1版.北京:科学出版社,2004:88-93
    [31]文贡坚,王润生.从航空遥感图像中自动提取主要道路、软件学报,2000(7):958-962
    [32]L.Bruzzone and D.Fernandez. An adaptive parcel-based technique forunsupervised change detection,Int.J.Remote.Sens.2001:pp.817-822
    [33]郎锐.数字图像处理学.第1版.北京:希望电子出版社,2002:235-345
    [34]边肇祺等.模式识别.北京:清华大学出版社,1999:1-256
    [35]汤国安等.遥感数字图像处理.北京:科学出版社,2004:66-71
    [36]S. Geman and D. Geman, Stochastic relaxation, "Gibbs Distribution and BayesianRestoration of Images", IEEE Trans. Pattern Recognition and Machine Intelligence,1984: pp234-256
    [37]Tuong Thuy Vu; Matsuoka, M.; Yamazaki, F., "LIDAR-based change detection ofbuildings in dense urban areas", in Proceedings, Geoscience and Remote SensingSymposium, 2004:pp.3413-3416
    [38]吴青.高分辨率遥感图像道路网提取技术研究.哈尔滨:哈尔滨工业大学,2006:30-60
    [39]肖志强,鲍光淑.基于GA的SAR图像中主干道路提取.中国图象图形学报,2004(1):94-97
    [40]冯学智等.基于数学形态学的道路遥感影像特征提取及网络分析[J].中国图象图形学报,2003(7):779-803
    [41]史文中等.从遥感影像提取道路特征的方法综述和展望.测绘学报,2001(3)
    [42]季方,鲍远律.基于外沿特征的栅格地图噪声去除算法[J].中国图象图形学报,2004(9):1063-1067
    [43]蒋斌.SAR图像道路提取方法研究.长沙:国防科学技术大学,2004:9-18
    [44]胡翔云.航空遥感影像线状地物与房屋的自动提取.武汉:武汉大学,2001:25-56
    [45]曹广真,金亚秋.基于水平集方法的多源遥感数据融合及城区道路提取.电子与信息学报,2007:1466-1467
    [46]朱长青等.基于结点扩展的高分辨率遥感影像道路提取.华中科技大学学报(自然科学版),2009(1):27-29
    [47]焦李成.高分辨率遥感图像中道路提取方法研究.西安.:西安电子科技大学,2009:20~45
    [48]王娜.地图道路、河流信息的识别与提取方法研究.贵州:贵州大学,2007:8-12
    [49]宫鹏等.高分辨率影像解译理论与应用方法中的一些研究问题[J].遥感学报,2006-10(1):2-3
    [50]王植,贺赛先.一种基于Canny理论的自适应边缘检测方法.中国图象图形学报,2004-8(10):958-960
    [51]陈劲松等.中分辨率遥感图像条带噪声的去除.遥感学报,2004-3(3):229~230
    [52]施海亮.基于Snake s模型的高分辨率遥感影像道路提取方法研究.南京:河海大学,2007:20-45
    [53]吴军,张万昌MODIS影像条带噪声去除的自相关插值法.遥感技术与应用,2006-03(3):254-256
    [54]洪日昌等.低分辨率遥感影像中道路的全自动提取方法研究.遥感学报,2008(1):37-39
    [55]朱长青等.基于结点扩展的高分辨率遥感影像道路提取.华中科技大学学报(自然科学版),2009-01(1):61-73
    [56]王莉莉等.基于D—S证据理论的城市航拍道路提取方法.软件学报,2005-9(2):1536-1538
    [57]五丰,秦其明.基于知识的卫星数字图像公路信息提取研究.北京大学学报,1998-1(1)
    [58]李燕.基于模型约束的遥感影像道路智能提取技术研究.郑州:中国人民解放军信息工程大学,2003:19-45
    [59]黄明华.航片图像纠正方法分析.安徽理工大学学报(自然科学版),2004(3):15-17
    [60]洪日昌等.低分辨率遥感影像中道路的全自动提取方法研究.遥感学报,2008(1):37-40
    [61]刘锦辉.图像增强方法的研究以及应用.湖南:湖南师范大学,2009:12~24
    [62]贾承丽.S A R图像自动道路提取[J].中国图象图形学报,2005-10(10):1218-1220
    [63]刘炜.基于概率模型的高分辨率卫星图像建筑物识别及变化检测.合肥:中国科学院,2005:37-51
    [64]余长慧等.基于标点随机过程的遥感影像道路提取.武汉大学学报,2006(1):60-61
    [65]陈芒.图像中道路网络自动识别的研究.上海:上海交通大学,2006:16-72

© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号

地址:北京市海淀区学院路29号 邮编:100083

电话:办公室:(+86 10)66554848;文献借阅、咨询服务、科技查新:66554700