It is possible to extract the lines and points 
from a photogrammetric stereomodel, if they are 
distinct enough. The problem is to define an 
objective criterion for detecting break lines and 
points. In this context, however, a method based 
on the concept of profile analysis by applying the 
second difference criterion, is used. 
The second difference in height of a triplet of 
points (V?h) is compared  vith a certain 
preselected threshold value, in case (V2h) is 
greater than the threshold, then the point belongs 
to the skeleton (I), otherwise, as filling (IT) 
information. 
Hence the total terrain relief information (T) is 
composed of the skeleton (I) and filling (II) 
T=201 (1) 
2.4.1 Classification of skeleton information 
for optimum sampling (I) The skeleton I can be 
classified (Charif, 1991) according to the: 
- Genetics 
. natural feature (IN) 
. man-made objects (Z0) 
iz INA I0 (2) 
- Geometric entities: 
. Lines (L): 
- Distinct break-lines (BL): Ridge lines (DR), 
Drainage lines (DD), Convex (DV), Concave (DC) 
BL = DR ADD A DV A DC (3) 
. Auxiliary (non distinct) lines (AL) Maxima 
(AX), Minima (AN), Others (AO) 
AL = AX A AN A AO (4) 
- peripheral lines (PL): Water (PW), Clouds 
(PC), other (PO) 
PL = PV A PC A PO (5) 
Thus, all lines together: 
L = BL AAL A PL (6) 
. Points (P): 
- Distinct break-points (BP): Peaks (DK), Pits 
(DT), Pass (DS), Convex (DE), Concave (DA) 
BP - DK A DT A DS A DE A DA (7) 
. Auxiliary (non distinct) points (AP): 
(AK), Pits (AT), Pass (AS), 
Concave (AA) 
Peaks 
Convex (AE), 
AP = AK A AT A AS A AE A AA (8) 
Thus, all points together: 
P = BP A AP (9) 
Moreover the skeleton (E) information can be 
differentiated further, according to the 
hierarchy of the information, to: 
Z = n * Z, + L4 (10) 
where I, is primary skeleton information, 
L is secondary skeleton information, etc ... 
2.4.2 Classification of the filling information 
The filling information (I) represents the 
80 
terrain relief other than the skeleton X. I is 
composed of incomplete regular grids of different 
densities. 
The classification is inherent in the grids of the 
successive sampling runs: 
IL = I, + I * I, + IL + n (11) 
A possible quantitative criterion for classifica- 
tion is the relationship between (ZI) and (IT) 
information. 
A) Natural terrain: 
1. Smooth terrain, where Lat = 0 
No of BL / unit area = 0.0 
2. Slightly rough terrain, vhere Znat « II 
No of BL / unit area - 0.0 to 0.02 
3. Moderately rough terrain, where lat « II 
No of BL / unit area = 0.02 to 0.05 
Iv 
= 
4. Very rough terrain, vhere Znat 
No of BL / unit area = 0.05 to 1.0 
B) Urban, industrial, rural terrain: 
1. Smooth terrain, where Z = 0 
art 
2. Slightly rough terrain, vhere La « II 
rt 
3. Moderately rough terrain, where i art « II 
4. Very rough terrain, where Y art 2m 
Sampling real terrain feature should be assessed 
using some quality assessment measures, the latter 
needs to be studied in detail. 
3. QUALITY MEASURES 
DTM is meant for various applications, obviously 
the quality of DTM varies according to intended 
application. The quality of DTM intended for 
irrigation or large scale application would be 
different than that intended for small scale 
application. The objectives of this study are to 
define the quality assessment model for DTM, and 
study the relationship between the terrain 
classification, sampling procedure, and quality 
assessment. 
The quality assessment of DTM is differentiated 
according to the performance (accuracy, fidelity), 
reliability, and efficiency (Charif, M., 1991). 
3.1 Performance 
As it was stated earlier the performance is one of 
the main criteria influencing the estimation of 
the quality of DTM products. Performance was 
differentiated further according to completeness 
of X information, Accuracy of X and Il information, 
and the fidelity of I and I information. 
3.1.1 Accuracy In Composite Sampling, the 
terrain relief is represented by the X and I 
sub-sets, consequently, the accuracy estimation 
should be differentiated according to; 
- The standard error c, of modelling by the Z set. 
This can be differentiated further according to 
can 
accura 
. Comp: 
comp] 
samp] 
on: 
. The 
tel 
Th» 
. Accu] 
on: 
« Im: 
. Pre 
. Ope 
. San 
The 
deper 
. Api 
« Gr) 
« PO 
. Int 
Because 
to the 
the Il-< 
modelli 
3.2 Sot 
The é 
influer 
+ Errol 
menti 
. Measi 
the 
mode! 
the : 
Assumir 
is the 
samplec 
8; (*) s 
In phot 
partly 
latter 
of ur 
distril 
varianc 
Assumin 
indepen 
the err 
where 
interpo