International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B4, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
between topographic features and background, it took nearly 
the same time to vectorize linear features by using our method 
and MapGIS. But for map images with low contrast and low 
SNR, our method is obviously efficient. It took about 9 minutes 
to vectorize counter lines in Figure 11 by using MapGIS, 
apparently slower than our method. Tracking errors (as shown 
in the white circle marked areas in Figure 13) often occur in the 
process of vectorization, and more human interventions are 
needed to handle these problems. 
  
Figure 13. Vectorization result of Figure 11 by MAPGIS 
( locally enlarged) 
From the experiments, the proposed method demonstrates the 
following advantages: 
(1) Most of the work of line tracking can be finished 
automatically while human operators only need to give the 
starting point and the directional point. Some kinds of manual 
interventions are allowed in case automatic tracking fails, 
which makes line tracking under human control, and provides 
the ability to correct data immediately if required. 
(2) Linear features can be tracked accurately along the 
centerline after image segmentation and thinning. This can 
avoid deviation from the centerline using only colour distance 
to determine the tracking point. 
(3) The sliding window is updated continuously and the 
segmentation result in each window depends on the grey-level 
distribution and spatial relationship of pixels in current window. 
This makes line tracking adapt to colour variations in light and 
shade areas. 
S. CONCLUSION 
This paper presents a method for linear feature vectorization 
directly in scanned colour maps. It has been proved to be valid 
and practical, especially for those maps with forest tints and 
relief shadings. The process of sliding window creation, 
adaptive image segmentation, thinning and sequential tracking 
can be used as general steps for colour map vectorization. 
Future improvement mainly focuses on automatic tracking 
across the intersections of different linear features so as to 
reduce human intervention and improve the speed of 
vectorization. 
References 
Cheng Qimin, Yang Chongjun, Shao Zhenfeng, 2003. 
Automatic Color Separation and Processing of Scanned Color 
Contour Maps. Journal of Geomatics, 28(5), pp. 4-6 (in Chinese) 
Chen Zeyu, Qi Feihu, Chen Gang, 2000. Human Face Detection 
Using Color Information. Journal of Shanghai Jiaotong 
University, 34(6), pp. 793-795 (in Chinese) 
Feng Yucai, Song Enmin, 1996. An Algorithm to gather colors 
from image of color maps. Journal of Software, 7(8), pp. 466- 
470 (in Chinese) 
Ford A, Roberts A, 1998. “Colour Space Conversions”. 
http://www.poynton.com/PDFs/coloureq.pdf (16 April. 2012) 
Fu Zhongliang, 2001. The Threshold Selection Method Based 
on the Measure of Difference between Images. Journal of 
Computer Research and Development, 38(5), pp. 563-567 (in 
Chinese) 
Guo Zihua, Zhuang Zhenquan, 1998. Map Vectorization 
Algorithm Based on Gas. Journal of China University of 
Science and Technology, 28(4), pp. 476-481 (in Chinese) 
Huang Zhili, Fan Yangyu, Hao Chongyang, 2005. Interactive 
Vectorization for Chromatic Scan Map. Computer Application, 
25(3), pp. 577-579 (in Chinese) 
Lin Hua, Wang Zhengguang, 1994. The Application of the 
ART Network in the Color Classification of Maps. Journal of 
Xidian University, 21(4), pp. 419-425 (in Chinese) 
Su Haihua, Wang Ruibin, Zhuang Zhenquan, 1999. A Colour 
Segmentation Algorithm Applied in Maps with Its 
Implementation. Journal of Computer-aided Design 
&Computer Graphics, 11(5), pp. 395-398 (in Chinese) 
Wagstaff K, Cardie C, Rogers S, et al, 2001. Constrained K- 
means Clustering with Background Knowledge. In: 
Proceedings of the Eighteenth International Conference on 
Machine Learning, Massachusetts, USA, pp. 577-584 
Wu Xincai, Wang Xiaorui, 1998. The Method of Interactive 
Vectorization Based on Color Map. Mini-Micro Systems, 19(5), 
pp. 36-39 (in Chinese) 
Zeng Yingsheng, Feng Jinguang, He Hangen, 2004. A 
Categorizing Algorithm in the Recognition of Contour Line. 
Computer Engineering and Application, 40(3), pp. 56-57+62 
(in Chinese) 
Zheng Huali, Zhou Xianzhong,Wang Jianyu, 2003. Research 
and Implementation of Automatic Color Segmentation 
Algorithm for Scanned Color Maps. Journal of Computer-aided 
Design &Computer Graphics, 15(1), pp. 29-33 (in Chinese) 
Acknowledgements 
The authors would like to thank Zondy Cyber Group Co., LTD 
for providing MAPGIS K9 to do the experiments. 
104 
KE 
AB 
pro 
sm 
hur 
con 
tim 
cpu 
clu 
tar, 
of« 
pro 
Ma 
intc