Full text: Mapping without the sun

186 
Dark car 
Figure 9. Different reflectance 
(3) Background confusing 
Some cars’ background is very confusing. Sometimes the 
background is detected as vehicle. 
Figure 10. Confusing background 
4. DISCUSSION 
The accuracy of moving vehicle detection is based on the 
techniques of satellite sensor model refinement, image 
resolution, accuracy of vehicle image coordinates, accuracy of 
satellite time interval, and DEM accuracy. The accuracy of 
satellite time interval and image resolution is related to satellite 
equipment. They can be considered as constants. 
Because the time interval of panchromatic image and multi- 
spectral image is very small, which means the intersection 
angle of the PAN and MS images is also very small, so we can 
not calculate vehicle ground coordinates just based on its image 
coordinates. DEM is necessary and its affection to the ground 
coordinates calculation should be limited in a reasonable range. 
DEM accuracy is a very complicated difficulty. Many research 
scientists are focusing on this topic. But in this research it is not 
focus. 
So the last aspect that can be improved is the accuracy of 
vehicle image coordinates. In our experiment, we used the 
mean gray value as threshold to select vehicle pixels, so as to 
calculate vehicle’s central position. But because the vehicles 
can have different color, they can have different reflectance. 
Some vehicles have more than one color, such as half bright, 
half dark, or top bright, the others dark. Therefore, for these 
vehicles, taking the mean gray value as threshold to select 
vehicle pixels is not very suitable. On the other hand, the 
vehicle is so small on the satellite imagery, so the color 
information is very limited. This is a problem we should pay 
attention to in the future. 
From the experiment, we noticed that the vehicle’s color, 
background, and their relative position all can affect the 
detection and later the accuracy. 
5. CONCLUSION 
We have presented the whole procedure of extracting vehicle 
information from Quickbird imagery based on an automatic 
method. It includes several steps: image classification, vehicle 
detection, image matching, calculation of image position on the 
PAN image and the MS image, and calculation of velocity. The 
experiment result shows that this technique can detect moving 
vehicle and extract vehicle’s position, moving speed and 
moving direction. But we recognized there is still very potential 
for further improvement in the vehicle image coordinates 
calculation, so that to improve the accuracy of the vehicle 
information. As the satellite time interval is very small and 
vehicle’s moving distance during this short time is also very 
limited, so even a very small improvement in the vehicle image 
coordinates measurement, say 0.1 pixels, will give a very big 
contribution to the accuracy of moving vehicle detection. This 
also is our next focus. 
We noticed that classification and image segmentation may not 
be the best techniques to detect vehicle. The accuracy is limited 
by the poor spectral feature of QuickBird satellite images. 
Therefore, a new method to accurately detect the vehicle is 
urgent and necessary. 
REFERECES 
Castellano, G.; Boyce, J.; Sandler, M. 1999. CDWT optical 
flow applied to moving target detection. IEE Colloquium on 
Motion Analysis and Tracking (Ref. No. 1999/103), pp. 17/1 - 
17/6 
Dias, J.M.B.; Marques, P.A.C., 2003. Multiple moving target 
detection and trajectory estimation using a single SAR sensor. 
IEEE Transactions on Aerospace and Electronic Systems, 
Volume 39, Issue 2, pp. 604 - 624 
Hong-Bo Sun; Guo-Sui Liu; Hong Gu; Wei-Min Su, 2002. 
Application of the Fractional Fourier Transform to Moving 
Target Detection in Airborne SAR. IEEE Transactions on 
Aerospace and Electronic Systems, Volume 38, Issue 4, pp. 
1416-1424 
Jacek Grodecki and Gene Dial, 2003. Block Adjustment of 
High-Resolution Satellite Images Described by Rational 
Polynomials. Photogrammetric Engineering & Remote Sensing 
Vol. 69, No. l,pp. 59-68 
Kaichang Di, Ruijin Ma, and Rong Xing Li, 2003. Rational 
Functions and Potential for rigorous Sensor Model Recovery. 
Photogrammetric Engineering & Remote Sensing Vol. 69, No. 
1, pp. 33-41 
Liu, C.-M. Jen, C.-W., 1992. A parallel adaptive algorithm for 
moving target detection and its VLSI array realization. Signal 
Processing, IEEE Transactions on Acoustics, Speech, and 
Signal Processing, Volume: 40, Issue: 11, PP- 2841 - 2848
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.