International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 Interna 
Difference of points Difference of points (95 
: Difference of pointe Difference of points (95%) 
5 15 | is 
" l | } in 
| | : 
8 s 1 | i $ ; | 5 
| LE 5 Q. T i bh 1 JE 
4j xx ib 5 rd ; 
ul By a Py i is, 
um T i E MOM M NUM ! t gp ve, 
1000 a NEO 00 re ND SS. 
m 100 x E um 
0 Neo Wm 5 
"n U 
Dif Distibution & Stat Df Distribution & Stat (8559) 
1 + x * 7 17 , + ' 3 D4 Distribution & Stat Cf Oistribution £& Stat (35%) 
1 1 
09 0a! : 
0.9 1 0.9 1 
on ve} 89 
i 08 1 Ce 1 
O7} pat Vide: 18544 07} Ipgut Values TET36 08 
Walid Values: 15295 or parcem 61 1664 | Vakd Values: 14787 or percent (O 0446 07} Mput Values. 72528 1 07} Input Values. 70777 1 Input Va 
0.6} mean: 0.499722 06} mesn. -0.53723 Valid Vslues. 52100 or percert B5 EOS V'slid Vstues: £0338 or percent 85.2500 07 MA 
irimrer mean 7% 4) 49915 | durmerusd rnéan 2Ue 0 3792 UB mean 0 10158 4 OB} man DEE J : vas 
06! stáder 17332 05] std der. 1.7455 trimmed mean 2*6: 0, 11162 trimmed mean 2% 411051 un m 
mesn sbsolute devistion: 14137 | mean absolute deastion | 2229 DS} stddew 0.94183 | 4 0.5} std dev 0.932123 |} à 95 en 
0.4} rangs: 10.2492 D4} range. 10.2492 1 trien abili: dign 1158337 an abel: devin O BALA cap e = 
skewness: 0.013064 | skewness. 0.020944 HA range; 21 2077 | 1 Ua ge: 20.8213 | | i ii sous al 
0.3} kudeas 2h42 1 n3: kutlasn 7 454 4 uke | JAHRE | tame D BO THE | A rage 
i "T kunosis: 13.9609 } | 1 UE} kurtosis 142101 | | skewne 
02 ~~ 0.2} J f | | 03? kırtoras 
| 0 i) n? | d uy 
a1 x 0.4} \ ; 
: | | \ 1 had 01 | \ 
o / ‘ x / à 0.1 
10 A D 5 10 Yl 5 Tn 5 10 "n 2 s u 2 
ia 5 0 = 10 10 5 0 5 10 n--———m 
1 
Figure 5. Visualization and statistical analysis of the SE ani 
differences between the manually collected and : 
the artificially distorted DEM (prior to 
. 15 15 
correction). = ‘ 
10 4- : : wl fs. 
; | npa M b | í 
These values become much more elegant and impressive if ; y At | li | 57 4 
one considers the same statistics for the DEM prior to siu SS af ) i 
corrections (fig. 5). The improvement of the shape of the no A s 5 
. v . . . = ne A uo ERA = T 
error distribution is devastating. Values of the statistical addo ! ux A TH # TUR. " 
analysis also justify the fact that the uncorrected DEM is aM n ig 7 ^ um 154 
5 iu M o 1 17 Aet MED C rat ip pos. 
much worse than the corrected one; mean -0.54, SD 1.74, Ses vom es _ 100 m 
oh E M oet 000 
MAD 1.42 meters and RMS 1.82. Le, eu CT i 
4.2 Application of the method over automatically 
collected DEMS. Uhil Pastibutiun € ital Did Chstribulion £ Sti (5%) 
1 Y I 
A recent research in the National Technical University of 08 | os | IF. 
Athens, laboratory of photogrammetry, concerned with image 08 i 0.0 us 
matching in color images, has automatically created 24 eri nu Values xm Pa 1 0 M 
: SER E dr = ; ES Valid Values. 28081 er percent 1OO d laid Ns 
DEMs using different software (Vision's Softplotter, Erdas HE} mean 0 13314 bl 1 ve hat es 
a Gn : . : niim 08527 | ME. 
OrthoPro and Z/I SSK). In order to test algorithm's integrity 05 H Ma] odd de DRAGS | | 1 id Be 
A ; NP bsolute deyist-an. 0.44352 mean absolute deviation 0.35326 + eme 
over real data, two DEMS created with Softplotter have been 0A} me MON 04f range 0736 | | 25) adder 
~ : : : ^ kewness. -1.1322 | | shewness: 018347) sap anal 
selected for testing with the proposed algorithm. Softplotter a3] es emo | | n3] katie 20700. | (at range 2 
= | A 1 sever 
allows the user to decide whether he wants the collected 02 j| ü2 i | 9.28 kurtesis 
points to be in a regular spacing or in random positions, 01 fo 01 Pd gs 
producing respectively Digital Terrain Model (DTM) and ü ; ol + ed $i 
S > ENS ur 10 E j # Yd 1 1 ^ n 
Iriangulated Irregular Network (TIN), accordingly to gm 
software's parameters (from now on, both freely referred as ; E set = 
Sere gp pae e A ow on, bc m 5 = ed > Figure 6. Initial (upper) and corrected DEM (lower) 
s). Th al surfaces are same, basically due to the ; : 1T eur 
diff 9. qne iie ur e s Not same, N ica ; t s ne comparison with the manually collected TIN. Figure 
erent appr in the collection procedure rather than 5; Y LAM ot: : ; c 
e et | pt eat amt f Sete I | ume ra an.to Visualization and statistics of the DEM collected 
the final interpolation performe he last step. epYT M? Cot 
ae Hm. Polation pe ep at the i e with the “DTM” method on the Softplotter. 
Althoug 
for sta 
summar 
556