nbul 2004 
nd second 
oint 2 and 
  
lak 
HBT 7 
L0 
= Zu) 
dependent 
(1h) 
each line 
ently, two 
quired to 
cting, the 
tersection 
| result, at 
O recover 
-coplanar 
; of the 3- 
involved 
s of the 
captured 
n average 
y of 2.24 
and range 
ing to the 
d 0.15 m 
mages are 
Jsing two 
and 4 m- 
distance 
m using 
dataset 
7 digital 
61 
  
  
  
   
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
  
3.1 LIDAR as control in photogrammetric triangulation 
LIDAR - RC-10 
  
Straight line segments extracted in LIDAR datasets I1 and I2 
were used in separate experiments as the source of control 
information for the photogrammetric triangulation. Table 2 
summarizes the quality of the aligned photogrammetric model 
through check point analysis. The results for the I2 dataset in 
Table 2 demonstrated some overall improvement either in the 
mean or standard deviation. 
  
  
  
  
  
  
  
  
  
LIDAR set I1 LIDAR set 12 
# of control lines 80 70. 
# of check points 32 32 
AX (m) 0.75 (£0.51) 0.65 (£0.28) 
AY (m) -0.10 (£0.43) -0.15 (+£0.26) 
AZ (m) -0.75 (£0.36) -0.69 (£0.42) 
  
Table 2: Check point analysis for LIDAR-RC-10 datasets. 
LIDAR-SONY F717 
Similar to the previous RC-10 experiments, straight line 
segments extracted in the two LIDAR datasets (11 & 12) were 
used in separate experiments as the source of control 
information for the photogrammetric triangulation of the SONY 
F717 dataset. Table 3 summarizes the quality of the aligned 
photogrammetric model through check point analysis. Again, 
significant improvement is noticed in the y-coordinate while no 
improvement is seen in the other directions. 
  
LIDAR set IH LIDAR set 12 
  
  
  
  
  
  
  
  
  
# of control lines 72 72 
# of check points 52 32 
AX (m) 0.38 (£0.63) 0.42 (£0.70) 
AY (m) 0.35 (0.70) 0.20 (x0.67) 
AZ (m) -0.49 (£1.11) -0.51 (£1.12) 
  
Table 3: Check point analysis for LIDAR-SONY F717 datasets. 
Comparing the results for the RC-10 dataset in Table 2 with that 
for SONY F717 in Table 3, it is clearly noticeable from the 
standard deviation values that RC-10 has a closer fit to the 
involved check points. This should be of no surprise since the 
expected accuracies based on the height-base ratio in Table 1 
indicated such a trend. Inspecting Table 2 again for the 
improvement in results between Il and 12 datasets, RC-10 
shows better overall results when the LIDAR was interpolated 
with the 1.0 m pixel size using the nearest neighbour method 
(I2 set). This improvement can be attributed to the fact that the 
1.0 m pixel size is closer to the 2.24 point/m” LIDAR point 
density (equivalent to 0.7 m pixel size), making it a more 
realistic sampling size. Table 3 shows that almost no change 
occurred between 11 and I2 interpolation sets for the SONY 
F717. Larger pixel size and lower level of detail in the SONY 
F717 images contributed to the steadiness in results as 
compared to the RC-10. 
In another look at the results for RCIO in Table 2, the mean 
values of the differences clearly suggest the existence of biases 
especially in the X- and Z- directions. The fact that the mean 
values in the same table are well above the standard deviation 
values supports this finding. Table 3 also indicates some biases 
in the SONY F717 results, but the standard deviation of these 
values are larger than the mean values; hence no conclusion can 
be drawn about the existence of such biases in any of the 
directions. 
3.2 LIDAR lines in the absolute orientation of the 
photogrammetric model 
A separate photogrammetric model, using both point and 
straight line features, has been generated for RC-10 and SONY 
F717 imagery sets. The datum for each model was established 
using the coordinates of precisely surveyed ground control 
points. Hence the resulting photogrammetric models actually 
represent the real object space. LIDAR lines are then utilized as 
control information in an absolute orientation procedure to align 
the generated photogrammetric object space through a 3D 
similarity transformation. Assuming the LIDAR and ground 
control points used in the photogrammetric reconstruction both 
have the same reference frame, one can use the parameters of 
the transformation function directly to assess the quality of fit 
between the two datasets. 
LIDAR - RC-10 
Conjugate lines in photogrammetric and LIDAR datasets were 
identified and measured. There were eighty lines in 11, seventy 
nine in I2, and twenty three from manually identified and 
intersected planar patches. Table 4 lists the transformation 
function parameters between LIDAR and RC-10 models. 
  
  
Il 12 Patch intersection 
Scale 0.999526 +0.00033 1.000097 +0.00025 1.000050 |=0.00038 
X m) 0.62.1 20.141 056 | £0.11. 033 | +015 | 
Yr(m) 0.19 £0.15 0.04 +0.11 -0.11 £0.14 | 
Zr (m)| -0.98 0.07 | -1.07 | £0.05 -0.86 | +£0.08 
Q(°)| -0.003 |+0.014 | -0.004 | £0.010 | 0.029 | +0.029 
d (°y| 0.030 140.012 0.029 | +0.009 | 0.083 +0.017 
K()| -0.020 | +0.018 0.009 | =0.013 | -0.023 0.021 
  
  
  
  
  
  
  
  
  
  
  
  
  
Table 4: 3D similarity parameters between LIDAR and RC-10 
models. 
The shift values in Table 4 for all sets indicate the existence of 
biases between the LIDAR and RC-10 datasets, thus confirming 
the results drawn from the first approach. After thorough 
investigation, it was found that RC-10 control points were 
recorded with respect to SAD 69 reference frame prior to 1998. 
On the other side, LIDAR data was based on SAD 69 after 1998 
adjustments. Certain biases especially in the X- and Z-directions 
have been reported between the two versions. The overall 
normal vector between conjugate photogrammetric and LIDAR 
lines before and after absolute orientation is calculated and 
showed in Table 5. In this table, the best results were for linear 
features extracted using LIDAR planar patch intersection, 
followed by 12 and then I1 datasets. 
  
[ | I1 | I2 | Patch intersection | 
Before absolute orientation 
DX (m) | -0.26 | £1.00 | -0.24 | 0.54 | -023 | 2025 
DY (m)] -0.15 | £1.09 | -0.01 | £0.56 | 0.02 | 0.20 
[DZ(m)| 0.96 | 20.64 | 1.01 | £0.65 | 0.72 | 037 | 
After absolute orientation 
DX (m) | 0.03 | +0.96 | 0.04 | £0.52 | 0.005 | 40.13 
|DY (m) | -0.03 | £1.04 | 0.02 | 20.54 | -0.057 | 0.12 
DZ(m)| -0.02 | £0.45 | -0.06 | £0.46 | -0.115 | £0.41 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Table 5: Overall normal vector between conjugate 
photogrammetric (RC-10) and LIDAR lines before and 
after absolute orientation.