Morphology based an automated approach for editing raster digitized contour maps

Morphology based an automated approach for editing raster digitized contour maps

Chen Xiaoyong, Shunji Murai
Institute of Industrial Science, university of Tokyo, Japan

Michio Kawahara, Yoshinori Yajima, Kentaro Narigasawa, Doihara Takeshi
Atsugi Technical Centre, Asia Air Survey Co., LID., Japan

Abstract:
The substantive preprocessing invariably required by current method to convert scan digitized contour maps into a digital elevation data base severely limits the use of raster scanned for many practical applications. This paper presents and efficient "morphology based" approach to edit contour lines digitized from topographic maps. In contrast to the predominantly manual and semi-automatic techniques currently in use to scan and capture topographic data from hard-copy contour maps, the proposed method mathematical morphology theory to reconstruct the topological connections of contour liens in rasterized contour maps. The approach has been successfully tested on a 1:50000 scale digitizing Japanese topographic map by suing C language and X window system on SUN IV workstations.

Introduction
Extraction of DTM data from existing topographic contour maps is of great importance to many applications such as landscape analysis, land form classification, civil engineering planning and design, geographic information systems, etc. For this reason, this kind of research has been widely interested in recent years [Baker, et.al. 1981; Callaham, 1983; Greenlee, 1987; Sukit and Murai, 1988, 1990; Chen and Li, 1990; Yang, 1990; Lin, et.al. 1990]. But up to now automated raster-vector conversion and elevation labeling of raster contour lines also need many manual interruptions and editings, more efficient methods and software are needed to speed up this procedure[Lichtner, 1988; Chen, 1990].

Based on mathematical morphology, this paper presented several kinds of relation based full-automated raster contour segmentation and linking methods. The key to these methods is the use morphological parallel image processing operators to automated get the topological relations between broken contour line points. By using these topological relations, together with the criteria of direction and distance, we can easily realize automated editing of raster contours. For improving the reliability of editing long distance broken contour lines, we proposed a post-editing method based on the morphological boundary of local broken contour areas. This method will be very useful to automated processing contour maps in complicated areas, such as urban areas. The approach has been successfully tested on a 1:50000 scale digitizing Japanese topographic map by using C language and X window system on SUN IV workstations.

Basic Procedures of system
For automated generation DMT from scanned topographical contour maps, we need many processing, such as scanning, thinning, contour image segmentation, raster contour editing and linking, generation of structure lines and geomorphologic points, raster-vector convertion, and DTM generation. These procedures can be simply described by the following flow chart(Fig.1).

Fig.1. Program Flow Chart of the System

Fig.3. Sgmentation of Contour Image
This paper doesn't with elevation labelling and DTM interpolation, which can be found in (Sukit and Murai, 1988, 1990; Chen and Li, 1990; Sircar, 1991).

Thinning and Segmentation
Color and gray-value contour image can be obtained from topographical maps by canning and A/D conversion. The methods of color image segmentation and gray-value image theresholding can be found in (Sukit and Murai, 1990; Nagao, 1980), and we do'nt want discuss these problems in this paper. After these processing, we can get binary contour image. usually, the binary contour image is not one pixel thickness, and includes many kinds of noise, such as grid line, letters and symbols for identifying contour lines. Therefore, we should do the processing of thinning and image segmentation firstly.

1 Thinning processing
There are several kind of morphological thinning method [Serra, 1982; Chen, 1991]. Among them, only the Homotopic Sequential Thinning (HST) can preserve the connectivity of image and get one pixel thickness. But HST method is not very fast since keeping the image connectivity should detect boundary pixel carefully. Here, we proposed a new thinning method which can processing more fastly and get one line thickniss connected contour line. This thinning method is consist of two steps. At the first step, by using Opening and Closing operators, we can fast get connected contour images which are within three line thickeniss. At the second step, by processing at most two times HST, we can get one line thickeniss connected contour line. This new thinning method can be described as follows [Fig.2]:

Tab. 1 Structure Elevent (Li)

Tab. 2 Structure Elevent (Ei) and H

Tab. 3 Structure Elevent {Ci}

Fig. 2. New Thinning Method
2 Image Segmentation

Noise point set detecting: Let X be skeletonized contour image, Opening X by structure element set {Ci} [Tab.3]; we can get noise crossing point set X1;
Contour image segmentation: Using set Difference procetion X/X1, we can segment noise crossing points from contour image[Fig.3 (b)];
Segmentation of number "0': Dilation broken contour point set and tracing contour arcs within dilation aares; if the end points of tracing arc not include broken points Pi and connected points Ti, then delete this arc[Fig.3 (c)-(d)].
Broken point set and direction tetecting: Based on hit-or-Miss operation and by using structure elements{Ei} [Tab.2], we can get broken contour point set P. Using 5*5 window and broken point coordinates (x,y), we can get broken point direction as [Fig.3 (f)].
Generation of broken point local areas and their boundaries: Using conditional Closing and set Difference operation, we can get local broken point areas and their boundaries [Fig.3 (e)].

Fig. 4 Neighbour Point Direction Method

Fig. 5 Connected Based Line Method
Relation Based Linking of Broken contour Lines

1 Neighbour Point Direction Method
By tracing the boundary line of a broken contour point local area, we can get tow neighbour points for each broken contour pointy, and by calculating the direction difference Di between neighbour points, then we can sequentially link neighbour points which direction difference have the min ê8-Dic ê value [Fig.4].

2. Connected Base Line Method
By Conditional Dilation of the broken point local area, we can get a small area which include connected contour line, and trace this area boundary and contour arcs included in this area. According to the trace sequence of point Si, we can find liking points Px, Py indirectly[Fig.5].

3. Post Vector Contour Linking Method
If the distance between broken contour points are too large, general linking method can not bet good result. For improving the reliability of automated editing method, we proposed a post editing method as [Fig.6]. The key to the method is the use of the boundary of local broken contour area as the morphological map boundary, just as the contour map has been cut several "holes", but the topological relationship between contour lines has been not changed. So we can labeling these broken contours directly. The processing of "morphological boundary" is just same as the processing of map boundary lines. After labeling each contour line, we can easily linking broken concourse based on its elevation value.

Practical Example and Conclusions
The above approach has been successfully tested on a 1:50000 scale digitizing Japanese topographic map by using C language and X window system on SUN IV workstations. Some results of a part map are shown in Fig.7. Fig.7 (a) - (d) show the procedure of contour image thinning and segmentation; Fig.7 (e) shows the procedure of contour linking; Fig.7 (f)- (h) show the procedures of contour raster-vector convertion and labeling; Fig.7 (i)-(j) show the procedure of DTM interpolation.

From above presented approach and example, we can find that mathematical morphology is very useful for automated getting topological relationship in raster data processing. Relation based broken contour automated editing is more useful and high reliability. Morphological boundary based post editing method can be used for automated processing of complicated areas, such as urban areas.

Fig. 6 Post vector contour Linking Method

Fig. 7 Practical Example