International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
Fig.5. Association of the original image with the image 
filtered with highpass filter in Fourier space (by 
addition of two images) 
  
  
Fig.6. Image on the left - a part of the original image, 
image on the right - result of enhancement 
3. AERIAL TRIANGULATION 
Data of three blocks of aerial images with different features 
and two scales, which are usually used in Poland, is shown in 
  
  
  
  
  
  
  
  
  
  
  
  
  
table |. 
Project name Piwniczna Bemowo | Echallens 
Type of terrain mountainous city flat 
Scale 1:28000 1:5000 1:5000 
Camera RC-20 LMK RC-10 
1000 
Focal length 150 300 150 
Cy[mm] 
No. of fiducial marks 4 8 4 
Film color color b/w 
Pixel size [um] 21 21 21 
GCP's natural natural |sygnalized 
No. of control points 5 6 5 
No. of check points 4 5 9 
No. of images 3 3 3 
RMS (a’priori) 30,50,30 | 25,25, 20 3,3,5 
GCT's (X. Y 73 [om] 
  
  
  
  
  
Table 1. Characteristic of examined images blocks 
To make automatic digital aerotriangulation Orthobase Pro of 
ERDAS software were used. This software realize this task 
using ABM method of image correlation - according to Grüns 
algorithm. 
Additional assumptions have been set up: 
— mean error of menstruation of fiducial marks on each 
picture after transformation can not exceed 10 um, 
— location of GCP's should provide regular coverage in 
the block, 
—  invariability of GCP's and independent check points 
(ICP) location after filtration. 
Image Analyst Z/I Imaging software was used for filtration and 
Orthobase Pro ERDAS Imagine. 
Following filtration methods have been tested: 
= filtration in spatial domain with gradient filters, 
= filtration in spatial domain with adaptive filter, 
= filtration in frequency domain with highpass filter. 
For such kinds of filtration I have choosen following ways of 
original images enhancement: 
with a view to simplify record informing about type of 
enhancement at the end of above mentioned description I have 
placed a number characteristic for particular filtration. In 
combination with the first letter of the images block name (P- 
Piwniczna, B-Bemowo, E-Echallens) this 1s univocal 
information about the block and the enhancement type. The 
digit 1 was reserved for original images. 
In spatial domain: 
= Original image enhanced by addition a result of 
filtration of original image using Sobel’s mask {9}, 
= Original image enhanced by addition a result of 
filtration of original image using Prewitt’s mask {3}, 
= original image enhanced by addition a result of 
filtration of original image using highpass filtrer {10}, 
« original image filtered with adaptive filter {6}, 
= original image filtered with adaptive filter with 
equalization of histogram {7}, 
« original image with equalization of histogram (5j. 
In frequency domain: 
= original image enhanced by addition a result of 
filtration of original image using highpass filtrer with 
ideal window {8}, 
= original image enhanced by addition a result of 
filtration of original image using highpass filtrer with 
Butterworth's window {2}, 
" original image filtered with highpass filter with 
Butterworth's window (4j. 
In spatial and frequency domain: 
* original image filtered with adaptive filter (in object 
space) and enhanced by addition a result of filtration 
of original image using highpass filter with 
Butterworth’s window (in frequency domain) {11}. 
The best results of automatic digital aerial triangulation are 
shown in tables 2, 3, 4. 
Internat 
  
  
Oo 
X, 
GC! 
  
check [ 
No. of 
Xo 
Yo t 
© 
  
Go 
R 
X, 
GCI 
R 
X; 
| check | 
No. of 
R 
Xo 
Yo [1 
  
« 
( 
  
Xo 
ut [ 
Lo 
o 
p } [ 
K 
  
Table 
Bingo-F 
have a 
images 
original 
and the