GISdevelopment.net ---> AARS ---> ACRS 1990 ---> Poster Session

A Topographic contour lines reading system

Lin Zongjian, Lu Jian & Yan Li
Wuhan Technician University of
Surveying and mapping China


Abstract
A semi automatic system for reading contour lines from topographic maps is proposed. Based on the map sheet image taking with CCD camera, a sequence of processing, such a line following, geometric rectification, mosaicking and DEM interpolation, are realized on this system.

Introduction
Topographic contour digitizing is heavy task in generating GIS data base. There are three ways to digitize map contours. The first is manual observing point height by overlapping transparent grid on the map sheet (Fig.1). The second is by using digitizing table, which is common on the market. Though the above mentioned two methods are popularly accepted, the efficiency is not very satisfying, and the accuracy depends on the operation. In the case of long time work, the accuracy are frequently going down due to the fatique of operator's eyes. The third way is automatically digitizing with scanner. However, fully automatic system is not functional well due to the complex drawing on the map and the limit capability of artificial intelligence at present.

A semi automatic system is proposed in this paper, attempting to make an optimal combination of automation and manual operation.

System components
The system components are shown in Fig. 2. Treasons of using CCD camera are: a) Geometric stability (can be calibrated) [1], b) high sensitivity with light, c) less volume, d) easy to change the image scale.


Fig.1 Contour lines reading


Fig.2 system components


Line following
  1. The Procedures of Line Following
    Topographic map is so complex with heavy load of geographic information. It is difficult to realize fully automatic map reading at the present time. However appropriately applying pattern recognition technology to get the accuracy and efficiency higher than manual operation is possible. In order to suit different topographic type two procedures for contour lien following were designed:

    • Automatic lien detection aided by a few scattered manual pointing.
      Manual pointing is needed only at obvious bend of contours line (Fig. 3). Off-course less than a half internal of contour is accepted and the deviation will be corrected automatically.
    • Automatic line following with manual directing at interrupted point
      At the point where automatic processing fails, the system generates a sound to ask for interactive operation. Manual pointing can also be appropriated.

  2. Line following with Multi-criterion
    There are many algorithms to be used for lien following. One of them is shown in Fig. 4. The central pixel in the image window indicated a point on a contour lien has just been detected. In order to find the next point, sums of 10 pixels on every bar of the 8 directions are calculated and compared. The minimum one is taken as the advanced direction. Though this criterion results are right in most case, mistakes are frequently found. Other algorithms shall be used, such as circuit detection, Laplacians criterion, Hough Transfer and different mistakes may be found. Based on facts that different criterion has different advantage points (3) (4), a new method called multi criterion line following is applied. The principle of this method can be described as follows:

    • Any confirm should be through more than two criteria.
    • The criterion which is successful at last point will be first applied for the present point. When a criterion fails, another will function automatically. The manual interactive operation is arranged at last of the queue.
    • The system s integrated by a series of algorithms, in which, any new algorithms is easy to be accepted for system updating.

Fig.3 Manual pointing


Fig.4 Line following


Geometric processing
  1. Conversion of Rater to Vector
    Line following is carried out in raster data structure overlapping the original image. The raster data are converted into vector form to be easy to geometric rectification and plotting.

  2. Geometric Rectification
    Because the following reasons, geometric rectification should be done for the results of contour of line following.

    • The orientation of the CCD camera is unknown. That will be solved by using 4 grid (e.g. kilometer grids) points as control points.
    • Geometric distortion of CCD camera needs to be corrected, which can be together with the geometric rectification.

  3. Mosaicking
    Because of the limited size of CCD array, a map sheet needs to be imaged patch by patch to obtain a satisfactory resolution. Therefore the rectified data should be mosaicked from patchs. To simplify the procedure, kilometer grids are used as the patch edges. All patchs are rectified into a common co-ordinate system.
Experiment and conclusion
A normal pointed topographic map has been chosen for the experiment. A patch of the results is shown in Fog. 5 in which the detected lines are completed overlapping on the original contour lines. The maximum deviation is less than 0.1mm, because the pixel is 0.2mm. The results can be displayed in both raster and vector form. Fig.6 is a Raster displaying.


Fig.5 Overlap display


Fig.6 Contour lines


References
  • T. Luhman etc. Proceeding on fast processing of photogrammetric data, ISPRS (1987) 35.
  • Y. Shirai, Three dimensional computer vision, Springer verlag (1987) 32
  • Lin Zongjian, Acta Geodetica et Cartogra-Phica Sinica, 14-2 (1985) 111
  • Lin Zongjian, Journal of Wuhan Technical university o surveying and Mapping 13-4 (1988) 37
  • Wang Zhizhuo, Principles of Photogrammetry, Publishing House of Surveying and Mapping Beijing.