The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part B3b. Beijing 2008 
Get the statistics for each road to determine the length and 
location. According to the given threshold (T2) of the road 
length to judge the line is whether the road edge line. Removed 
the one which is less than the threshold value, retain a certain 
length of the straight line or curve. We think it as the road that 
we extracted. Based on this perceptual grouping principle, not 
only the edge of the single-shape road can also be detected in 
the image as well as the crossing parts and the comers at last. 
According to the characteristics of four roads, shape and grey 
value of the extracting region, the road model and the 
knowledge of the road, road region from the initial image is 
extracted, that is road edge lines. 
3.6 Extracted Road in Original Image 
With coordinate information, draw the red line on the original 
image, which stands for the edge line and centre line of the road. 
This can indicate the right location of the road information 
clearly. 
4. EXPERIMENT 
4.1 Experimental Data Resource 
Data source in experiment comes from Quick Bird remote 
sensing image with 0.6m resolution, which shows the region of 
Lujiazui in Shanghai. Crop four sub-images(Figure 5-8) with 
four different types of road information on them. Test this 
improved method approached in the paper. All the algorithms 
are implemented in the VC 6.0 environment. 
Experimental image Figure 2-5 
4.2 Result of Edge Detection 
Edge detection image after using algorithm mentioned in 3.1. 
Figure 9. Edge Detection to Figure 5. 
Figure 10. Edge detection to Figure 6. 
Figure 11. Edge to Figure 7 Figure 12. Edge to Figure 8 
4.3 Experimental Result 
Figure 13 to Figure 16 are the final images after the work of 
extraction. Red lines refer to the edge lines and the centrelines 
of the road, overlapping into the original image. You can see 
the road information more conveniently. 
Single- 
General 
Cruciform 
Elevate 
Shape 
Crossing 
Cross 
Urban 
Road 
Road 
Highways 
T2 
(pixel) 
100 
50 
60 
35 
Table 1 Set the value of T2 
From the table 1, we can see that different shape of road has 
different the value of T2.In general, the more complicate shape 
of road, the less the value of T2.