APHY 
RICHE 
P13; 
kelyhood, 
g the land, 
ihalanobis; 
> semi arid 
traditional 
> shall call 
pplied in 
stablished 
R, 1994) 
ly to the 
d we will 
ot find a 
  
atural 
;d in solid 
in table on 
nation of 
synthetic 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
character of the pixel which integrates the topographic 
conditions, moisture, climatic, ecological of the observed 
target. This mixing of information is at the same time the 
«strength» of the spatial image and paradoxically, it could 
constitute a mask which would prevent the direct access 
to the object «soil» searched. This mask is different every 
time and its identification or interpretation would come 
back to unveil the anomaly. 
    
  
   
  
  
  
Légende 
3] Alluvial 
Calcimagnesic 
Colluvial 
Hydromorphic 
Isohumic 
[] 
[] 
[3 
E 
B Lithosol 
[] 
E] 
ü 
B 
  
Regosol 
B] Complex unite 
Steet 
Township 
   
Vers Dun o Bong 
Vers Jin BR 
Figure 3. Pedological map of the township of KSAR SBAHI 
and BERRICHE 
      
Samples choice of the soils 
classifies 
Training phase 
and 
Statistical analysis 
Y. 
Automatic classification Neximmn 1 kelihood: minimum distance: 
pornllelepiped: Mibalabonis and Spectral Angle Nap 
Results analyzes 
  
  
Anomalies 
Interprétation 
Interprétation 
Jormalisation of 
sen 
hypothesis 
  
        
Figure 4. General organization chart of the approach 
The powerful of the anomaly contributes to the 
understanding of factors which influence the spectral 
signatures considerably and to the identification of limits 
101 
of remote sensing when it comes to spectral 
characterization of objects on the ground (figure 5). 
  
  
BM Unclassified Courbes spectroles des sols 
Minite Complexe : 1 T 7 
v 0.40 3 
Wiithosol 2 Ë ] 
5 zn À 
n ; a i 
sol aluvin: N x 4 
OÖ nznr 4 
Wisohumique g 03r > 1 
hydromorphe E [ d = f Ne” + 
Mnuages d 0.25 ee j X. EE OR 1 
Bo x Zi ] 
céréales isis Lr 1 
urbain % 0.20} 4. d 
s t 7 N 
Byin d’'alep 5 E uertit " 
= e 
= 
5 
[S 
1 2 3 | 4 
Les bondes TM du 25/09/89 
  
Figure 5. spectrals signatures characterization of 15" training 
regions of different objects. 
Six algorithms of classification were exploited: the 
Maximum Likelihood, the minimum distance, 
MAHALANOBIS, the method parallelepiped and the 
approach of SAM « Spectral Mapper Angle». The choice 
of the samples- image is made on the basis of the 
pedological map in the BERRICHE plain (BNEDER, 
1994), where 13 range of samples (representative 10 
classes of soils) were selected. 
3. RESULTS AND DISCUSSIONS 
We confronted the results of different classifications 
with the pedological digitised map before, by the research 
of the pixels well classified with regard to the map and 
those that were badly classified and/ or unclassified 
(Figure 6). 
399()000 
3985000 
B lnon classer 
MH BUnite Complexe 
3 25 Wiithosol 
«+ Msol alluvial 
“ Misohumiqué 
Gé Mhydromorphe 
i 
3980990 
  
Figure 6a: Results of classifications by Maximum Likelihood method 
(northern part) of the September 1989 image