栅格地图中地理对象的矢量化研究及系统实现
|
摘要
地理信息的提取和识别是地理信息系统(GIS)、全球定位系统(GPS)发展的基础和迫切需要。栅格地图矢量化已经成为获取地理信息的主要途径之一。本文在研究、分析目前具有代表性的扫描图像矢量化方法的基础上,针对栅格图像的特点,用Visual C++6.0编程语言实现了一个地图矢量化系统RasToVec。
本文采用目前比较流行的基于细化的矢量化方法作为整体设计思想。首先根据地图不同地理对象灰度级不同的特点,将地图中的文字等标注从地图中分离,然后用最大类间方差阈值分割法将地图进行二值化。对二值化后的地图运用数学形态学进行噪声滤除和轮廓提取。然后对图像进行细化,细化时采用一种基于标记的保留节点域的细化方法,该方法避免了传统细化方法中节点变形从而改变图形拓扑结构的问题。
对细化后的图像矢量化时,本文根据地图的整体拓扑特征,先将地图中的节点域和连通弧段提取出来,然后利用遗传算法将连通弧段进行矢量化得到连通矢量弧段,根据最长延伸原则合并矢量弧段,确定节点,并得到最终的矢量段。矢量化算法除了采用遗传算法外,还采用了道格拉斯算法,最后对这两种算法的效果和效率进行了比较。在细化及连通弧段编码过程中都采用边处理边擦除象素的方法,有效避免象素的重复处理,降低图像的复杂度,并提高了矢量化的速度。
目前矢量化研究大多是在工程图领域进行的,本文提出的算法是根据地理信息系统的要求设计的,在保持地图拓扑结构及矢量化准确性方面取得很好的效果,并实现了与GIS软件MapInfo的接口,在地理信息系统领域有一定的理论意义和应用价值。
Extraction and recognition of the geography information is the basis of geographic information system(GIS) and global position system(GPS). The vectorization of grip maps has become one of the main methods of getting digital information. After studying some typical map vectorization methods, the author realizes a map vectorization software, RasToVec, with Visual C++ programming language to recognize and capture the geographic elements in the grid maps.
The current prevalent vectorization method based on thinning algorithm is adopted in the paper. Firstly, words labels in the scanned gray map are separated from the map as its gray rank is different from the other objects' in the map. Then the gray map is transformed into binary map by maximal threshold variance method. Noises in the binary map are filtered and geographic elements' edges are extracted by morphology. In the process of thinning the image, a new thinning method based on labeling is put forward. Compared with the traditional thinning algorithm, this method avoids the node distortion and map topology changing.
In the process of vectorization, the node area and connected segment are abstracted according to the whole topology of map. The connected segment is transformed into its vector form by genetic algorithm(GA). The longest expended principle is used to combine the vector segment, and then the nodes are ensured and the last vector list is got. Besides GA, Douglas-poiker method is used in the process of vectorization and the comparison of these two methods in efficiency and precision is presented in the end. The erasing pixels process after thinning and segments coding avoids re-recognition and complexity, so the speed is improved.
At present the research on vectorization is mainly among the engineering images. The method proposed in the paper satisfies the demand of GIS and realizes the interface with software MapInfo. The experimental results show that this method can reestablish the map's topology and get a good performance in precision and efficiency. This paper presents certain significance and application value in the field of GIS.
本文采用目前比较流行的基于细化的矢量化方法作为整体设计思想。首先根据地图不同地理对象灰度级不同的特点,将地图中的文字等标注从地图中分离,然后用最大类间方差阈值分割法将地图进行二值化。对二值化后的地图运用数学形态学进行噪声滤除和轮廓提取。然后对图像进行细化,细化时采用一种基于标记的保留节点域的细化方法,该方法避免了传统细化方法中节点变形从而改变图形拓扑结构的问题。
对细化后的图像矢量化时,本文根据地图的整体拓扑特征,先将地图中的节点域和连通弧段提取出来,然后利用遗传算法将连通弧段进行矢量化得到连通矢量弧段,根据最长延伸原则合并矢量弧段,确定节点,并得到最终的矢量段。矢量化算法除了采用遗传算法外,还采用了道格拉斯算法,最后对这两种算法的效果和效率进行了比较。在细化及连通弧段编码过程中都采用边处理边擦除象素的方法,有效避免象素的重复处理,降低图像的复杂度,并提高了矢量化的速度。
目前矢量化研究大多是在工程图领域进行的,本文提出的算法是根据地理信息系统的要求设计的,在保持地图拓扑结构及矢量化准确性方面取得很好的效果,并实现了与GIS软件MapInfo的接口,在地理信息系统领域有一定的理论意义和应用价值。
Extraction and recognition of the geography information is the basis of geographic information system(GIS) and global position system(GPS). The vectorization of grip maps has become one of the main methods of getting digital information. After studying some typical map vectorization methods, the author realizes a map vectorization software, RasToVec, with Visual C++ programming language to recognize and capture the geographic elements in the grid maps.
The current prevalent vectorization method based on thinning algorithm is adopted in the paper. Firstly, words labels in the scanned gray map are separated from the map as its gray rank is different from the other objects' in the map. Then the gray map is transformed into binary map by maximal threshold variance method. Noises in the binary map are filtered and geographic elements' edges are extracted by morphology. In the process of thinning the image, a new thinning method based on labeling is put forward. Compared with the traditional thinning algorithm, this method avoids the node distortion and map topology changing.
In the process of vectorization, the node area and connected segment are abstracted according to the whole topology of map. The connected segment is transformed into its vector form by genetic algorithm(GA). The longest expended principle is used to combine the vector segment, and then the nodes are ensured and the last vector list is got. Besides GA, Douglas-poiker method is used in the process of vectorization and the comparison of these two methods in efficiency and precision is presented in the end. The erasing pixels process after thinning and segments coding avoids re-recognition and complexity, so the speed is improved.
At present the research on vectorization is mainly among the engineering images. The method proposed in the paper satisfies the demand of GIS and realizes the interface with software MapInfo. The experimental results show that this method can reestablish the map's topology and get a good performance in precision and efficiency. This paper presents certain significance and application value in the field of GIS.
引文
[1]张海荣.地理信息系统原理与应用[M].徐州:中国矿业大学出版社,2008.
[2]汤国安,刘学军,闾国年,等.地理信息系统教程[M].北京:高等教育出版社,2008.
[3]张超.地理信息系统应用教程[M].北京:科学出版社,2007.
[4]毛凯,龙毅.基于旋转预处理的地图符号识别方法[D].武汉:武汉大学,2005.
[5]李朝锋,曾生根,许磊.遥感图像智能处理[M].北京:电子工业出版社,2007.
[6]罗照,周晓光.地图扫描矢量化关键技术研究[D].长沙:中南大学,2006.
[7]郝向阳.地理信息识别与提取技术[M].北京:测绘出版社,2001.
[8]Armenakis C,Leduc F,Cyr I,et al.A comparative analysis of scanned maps and imagery for mapping applications[J].Photogrammetry & Remote sensing,2003,57:304-314.
[9]Mena J B,Malpica J A.An automatic method for road extraction in rural and semi-urban areas stating from high resolution satallite imagery[J].Pattern Recognition Letters,2005,26:1201-1220.
[10]罗喆帅,毛奇凰,葛艳.工程图纸矢量化软件的设计与实现[J].微机发展,2004,14(5):32-38.
[11]彭京亮.一种快速的扫描铅笔稿图细化算法[J].中国图象图形学报,2000,5(5):434-439.
[12]Yuncai Liu.An automation system generation of digital map data from pictorial map resources[J].Pattern Recognition,2002,35:1973-1987.
[13]贺继刚,杨晓伟,吴广潮,等.基于模板保留的快速并行细化算法[J].计算机应用与软件,2007,24(12):26-28.
[14]梅园,孙怀江,夏德深.一种基于改进后模板的图像快速细化算法[J].中国图像图形学报,2006,11(9):1306-1311.
[15]程东旭,秦新强,张大发,等.基于区域增长的轮廓线提取算法[J].西安理工大学学报,2005,21(4):413-417.
[16]李岩影,郭静珺.大比例尺地图扫描图像的整体矢量化方法研究[J].测绘通报,2003,8(4):20-22.
[17]王强,谭建荣.基于图形约束的工程图扫描图像直线整体识别方法[J].计算机学报,2002,17(8):561-569.
[18]杨容浩,余代俊.扫描地图自动矢量化若干细节问题的解决方法[J].四川测绘,2008,31(3):127-136.
[19]许锋,卢建刚,孙优贤.神经网络在图像处理中的应用[J].信息与控制,2003,32(4):344-351.
[20]李敏强,寇纪淞,林丹,等.遗传算法的基本理论与应用[M].北京:科学出版社,2002.
[21]王小平,曹立明.遗传算法理论、应用与软件实现[M].西安:西安交通大学出版社,2002.
[22]Prasad L,Skourikhine A N.Vectorized image segmentation via trixel agglomeration[J].Pattern Recognition,2006,39:501-514.
[23]赖有华.工程图自动识别中图像预处理的研究与实现[D].北京:中国农业大学,2006.
[24]张明君,高有行.基于位图的自动矢量化技术研究[D].西安:西安电子科技大学,2004.
[25]杜奇,向健勇,袁胼春.一种改进的最大类间方差法[J].红外技术,2003,25(5):33-36.
[26]Soille R.形态学图像分析原理与应用[M].王小鹏等译.北京:清华大学出版社,1989.
[27]王家隆,郭成安.一种改进的图像模板细化算法[J].中国图象图形学报,2004,9(3):297-301.
[28]包建军,樊菁.一种改进的图像模板细化算法[J].计算机辅助工程,2006,15(4):43-46.
[29]Mena J B.Automatic vectorization of segmented road networks by geometrical and topological analysis of high relolution binary images[J].Knowledge-Based Systems,2006,19:704-718.
[30]侯云花,杨德宏.光栅自动矢量化的算法及实现[J].中国西部科技,2007,9:43-46.
[31]Kenneth R.数字图像处理[M].朱志刚,石定机等译.北京:电子工业出版社,2002.
[32]Alexander K,Pasi F.Reduced-search dynamic programming for approximation of polygonal curves[J].Pattern Recognition Letters,2003,24:2243-2254.
[33]陶长武,蔡自兴.现代图像压缩编码技术[J].信息技术,2007,15:53-56.
[34]原岛博.图像信息压缩[M].薛培鼎,徐国鼐译.北京:科学出版社,2004.
[35]Kasturi R,Fernandez R,Amlani M L,et al.Map Data Processing in Geographic Information Syestems[J].IEEE Computer,2000,12:304-314.
[37]田玉敏,刘国景.光栅图形矢量化方法分析与评价[J].计算机应用研究,2002,19(3):6-8.
[38]胡晋山,康建荣.地图扫描数字化误差分析及控制[J].测绘科学,2005,30(2):90-91.
[39]蔡良伟,胡世曦.基于相似性遗传算法及其在JSP中的应用[J].深圳大学学报,2006,23(2):107-110.
[40]陶志,许宝栋.基于遗传算法的粗糙集知识约简方法[J].系统工程,2003,7(4):116-122.
[41]张巍,郑建彬.基于遗传算法实现签名曲线的匹配[J].武汉理工大学学报,2008,30(1):39-42.
[42]钟云海,郑海,王孝通.矢量图的统一表示框架[J].中国航海,2006,(4):20-22.
[43]陈建春.Visual C++开发GIS系统[M].北京:电子工业出版社,2000.
[44]向世明.Visual C++数字图像与图形处理[M].北京:电子工业出版社,2002.
[45]钱能.C++程序设计教程[M].北京:清华大学出版社,2005.
[46]任爱华.计算机图形学[M].北京:北京航空航天大学出版社,2005.
[2]汤国安,刘学军,闾国年,等.地理信息系统教程[M].北京:高等教育出版社,2008.
[3]张超.地理信息系统应用教程[M].北京:科学出版社,2007.
[4]毛凯,龙毅.基于旋转预处理的地图符号识别方法[D].武汉:武汉大学,2005.
[5]李朝锋,曾生根,许磊.遥感图像智能处理[M].北京:电子工业出版社,2007.
[6]罗照,周晓光.地图扫描矢量化关键技术研究[D].长沙:中南大学,2006.
[7]郝向阳.地理信息识别与提取技术[M].北京:测绘出版社,2001.
[8]Armenakis C,Leduc F,Cyr I,et al.A comparative analysis of scanned maps and imagery for mapping applications[J].Photogrammetry & Remote sensing,2003,57:304-314.
[9]Mena J B,Malpica J A.An automatic method for road extraction in rural and semi-urban areas stating from high resolution satallite imagery[J].Pattern Recognition Letters,2005,26:1201-1220.
[10]罗喆帅,毛奇凰,葛艳.工程图纸矢量化软件的设计与实现[J].微机发展,2004,14(5):32-38.
[11]彭京亮.一种快速的扫描铅笔稿图细化算法[J].中国图象图形学报,2000,5(5):434-439.
[12]Yuncai Liu.An automation system generation of digital map data from pictorial map resources[J].Pattern Recognition,2002,35:1973-1987.
[13]贺继刚,杨晓伟,吴广潮,等.基于模板保留的快速并行细化算法[J].计算机应用与软件,2007,24(12):26-28.
[14]梅园,孙怀江,夏德深.一种基于改进后模板的图像快速细化算法[J].中国图像图形学报,2006,11(9):1306-1311.
[15]程东旭,秦新强,张大发,等.基于区域增长的轮廓线提取算法[J].西安理工大学学报,2005,21(4):413-417.
[16]李岩影,郭静珺.大比例尺地图扫描图像的整体矢量化方法研究[J].测绘通报,2003,8(4):20-22.
[17]王强,谭建荣.基于图形约束的工程图扫描图像直线整体识别方法[J].计算机学报,2002,17(8):561-569.
[18]杨容浩,余代俊.扫描地图自动矢量化若干细节问题的解决方法[J].四川测绘,2008,31(3):127-136.
[19]许锋,卢建刚,孙优贤.神经网络在图像处理中的应用[J].信息与控制,2003,32(4):344-351.
[20]李敏强,寇纪淞,林丹,等.遗传算法的基本理论与应用[M].北京:科学出版社,2002.
[21]王小平,曹立明.遗传算法理论、应用与软件实现[M].西安:西安交通大学出版社,2002.
[22]Prasad L,Skourikhine A N.Vectorized image segmentation via trixel agglomeration[J].Pattern Recognition,2006,39:501-514.
[23]赖有华.工程图自动识别中图像预处理的研究与实现[D].北京:中国农业大学,2006.
[24]张明君,高有行.基于位图的自动矢量化技术研究[D].西安:西安电子科技大学,2004.
[25]杜奇,向健勇,袁胼春.一种改进的最大类间方差法[J].红外技术,2003,25(5):33-36.
[26]Soille R.形态学图像分析原理与应用[M].王小鹏等译.北京:清华大学出版社,1989.
[27]王家隆,郭成安.一种改进的图像模板细化算法[J].中国图象图形学报,2004,9(3):297-301.
[28]包建军,樊菁.一种改进的图像模板细化算法[J].计算机辅助工程,2006,15(4):43-46.
[29]Mena J B.Automatic vectorization of segmented road networks by geometrical and topological analysis of high relolution binary images[J].Knowledge-Based Systems,2006,19:704-718.
[30]侯云花,杨德宏.光栅自动矢量化的算法及实现[J].中国西部科技,2007,9:43-46.
[31]Kenneth R.数字图像处理[M].朱志刚,石定机等译.北京:电子工业出版社,2002.
[32]Alexander K,Pasi F.Reduced-search dynamic programming for approximation of polygonal curves[J].Pattern Recognition Letters,2003,24:2243-2254.
[33]陶长武,蔡自兴.现代图像压缩编码技术[J].信息技术,2007,15:53-56.
[34]原岛博.图像信息压缩[M].薛培鼎,徐国鼐译.北京:科学出版社,2004.
[35]Kasturi R,Fernandez R,Amlani M L,et al.Map Data Processing in Geographic Information Syestems[J].IEEE Computer,2000,12:304-314.
[37]田玉敏,刘国景.光栅图形矢量化方法分析与评价[J].计算机应用研究,2002,19(3):6-8.
[38]胡晋山,康建荣.地图扫描数字化误差分析及控制[J].测绘科学,2005,30(2):90-91.
[39]蔡良伟,胡世曦.基于相似性遗传算法及其在JSP中的应用[J].深圳大学学报,2006,23(2):107-110.
[40]陶志,许宝栋.基于遗传算法的粗糙集知识约简方法[J].系统工程,2003,7(4):116-122.
[41]张巍,郑建彬.基于遗传算法实现签名曲线的匹配[J].武汉理工大学学报,2008,30(1):39-42.
[42]钟云海,郑海,王孝通.矢量图的统一表示框架[J].中国航海,2006,(4):20-22.
[43]陈建春.Visual C++开发GIS系统[M].北京:电子工业出版社,2000.
[44]向世明.Visual C++数字图像与图形处理[M].北京:电子工业出版社,2002.
[45]钱能.C++程序设计教程[M].北京:清华大学出版社,2005.
[46]任爱华.计算机图形学[M].北京:北京航空航天大学出版社,2005.