tric 
ive 
wn 
our 
ive 
ain 
the 
lon 
  
each 
ould 
hical 
ould 
ation 
CPU 
and 
| this 
vidia 
GHz 
ance 
ation 
sults 
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 
  
  
  
  
  
  
  
  
Image Size GPU Time CPU Time CPU/GPU 
(pixel) (millisecond | (millisecond | (rate) 
) ) 
1024 x 1024 | 14,09 48,917 3,47 
2048 x 2048 | 75,994 190,626 2.51 
4096 x 4096 | 175,593 790,991 4,50 
  
Table 1. Calculation results with using GPU and CPU. 
S. DIRECT GEOREFERENCING WITH CUDA 
Direct Georeferencing is the direct determination of the position 
and orientation parameters of a sensor. It is an enabling 
technology for quantitative data acquisition and mapping 
applications where precise orientation and the position of the 
sensor are required. A direct georeferencing system provides 
the position and orientation of the sensor required to register the 
acquired data in geographic coordinates. In photogrammetry, 
direct georeferencing is used to produce measurement of the 
exterior orientation parameters for each image without use of 
ground control points or acrial triangulation. 
Direct georeferencing is based on colinearity equations to 
provide a relationship between pixel positions and 
corresponding positions on ground. The procedure of 
differential rectification is applied in combination with the 
forward projection (direct) method of orthoimage reprojection. 
This is based on the colinearity principle, which states that the 
projection center of a central perspective image, an object point, 
and its photographic image lies upon a straight line. 
Direct Georeferencing for aerial digital frame cameras consists 
of six stages and the data needed for each step are shown in 
Figure 8 (Kiraci, 2008). 
( Sat y 
A 
Image 
Coordinates 
  
  
UJ 
  
  
/ Parameters 
/ Inner 
ie 7 Orientation / 
LI lul 3 
Camera 
Coordinates 
  
  
  
7 
f Exterior Orientation Parameters // 
/ WGS84 Parameters 
Zz Atmospheric Parameters / 
e 
  
     
    
   
WGS84 
Cartesian 
Coordinates 
  
fe SN 
/ DEM / 
7 WGS84 
/ Parameters / 
Projection / 
Parameters / 
  
   
  
  
Eipsoida: 
Coordinates 
  
   
UTM 
Coordinates 
  
      
  
Origin / 
erre med, 
Orthophoto 
Coordinates 
   
PPT: Pixel by Pixel 
Transformation 
> 
— 
Figure 8. Direct Georeferencing algorithm. 
  
It is shown in figure that the procedure is really suitable for 
GPGPU and CUDA programming. Because of that the 
procedures must be done for each pixel of image. In every loop 
we must do pixel by pixel transformation. So all these 
calculation procedures parallelized and coded with using 
CUDA programming language. 
The program running results are; calculation GPU time 
1328234 millisecond, for the same procedure CPU time is 
6827282 millisecond. The discrepancy with GPU and CPU 
time shows that GPU 5.14 times faster than CPU (Table 2). In 
this example we used 4096x4096 pixel size image and Nvidia 
Geforce 8600M GT graphic card with Core 2 Duo 2.2GHz 
CPU. If the image size getting smaller, CPU performance 
increase. So with the very huge and repetitive calculation 
problems using GPU is more efficient and produces results 
rapidly. 
  
  
  
  
  
  
  
  
  
Image Size GPU Time CPU Time CPU/GPU 
(pixel) (millisecond | (millisecond | (rate) 
) ) 
1024x 1024 | 84675 324563 3,83 
2048 x 2048 | 342634 934547 2,73 
4096 x 4096 | 1328234 6827282 5,14 
  
Table 2. Calculation results with using GPU and CPU. 
6. RESULTS 
In this study covers how general purpose parallel programming 
and computational power of the GPUs and GPGPU method can 
be used in photogrammetric orthorectification applications 
especially direct georeferencing and projective rectification. 
These two methods coded with CUDA programming language. 
The results obtained are evaluated; the method is really suitable 
for image processing and photogrammetry especially if we do 
the same calculations to per image pixels. Also it is suitable for 
intensive calculation procedures. GPGPU and CUDA 
programming method make the calculation really fast. We can 
increase the number of applications which can be adapted to 
photogrammetry and image processing that require intensive 
computation and speed. 
In our study we didn’t make any optimization of the CUDA 
programming code. So in future studies we will focus on 
optimization of the coding procedures. Also our development 
software doesn’t have a user interface yet. We will make a user 
interface for this software. 
Today, with this method the images obtained through a variety 
of platforms, to be georeferenced correctly and quickly. 
Especially if it is important that real time processing of imagery 
airborne data for natural disasters, for rapid evaluating data 
coming from unmanned air vehicles (UAV) in military 
applications, for supporting rescue and security forces, and also 
for obtaining surveys in disaster scenarios or mass events an 
airborne real time image processing system is required. For this 
purpose we can use this GPGPU method for rectifying process 
of imagery data. 
159