OPTIMUM SAMPLING FOR DIGITAL TERRAIN MODELLING 
M. Charif 
ITC, Department of Geoinformatics, P.0. Box 6, Enschede, The Netherlands, ISPRS Commission III 
ABSTRACT: 
The investigation addresses three main issues: The first concerns a study of the terrain relief 
classification for the purpose of composite sampling. The second issue concerns quality assessment models 
that are used to assess the quality of representation of ideal geometric primitives and of real terrain 
relief representation. The. third issue concerns the performance of composite sampling when applied to 
real terrain relief. In order to test the rules drawn from idealised geometric primitives on real terrain 
relief, two different terrain models were selected. First the skeleton information was sampled 
selectively, then Composite sampling was carried out. Finally, the quality was assessed implementing the 
quality assessment model. The analysis of the test results provides a feedback for optimizing the 
procedure for Composite Sampling. 
KEY WORDS: Selective sampling, progressive sampling, skeleton information, filling information, composed 
transfer function, break line, peripheral line, rule base. 
  
  
  
INTRODUCTION 
SELECTIVE EXTRACTION AND SEGMENTATION 
Digital Terrain Model (DTM) is a digital (X) 
representation of the terrain relief serving for T 
various purposes such as automatic contour Y 
plotting, digital orthophoto production, 
calculation of earth-work volumes etc. SELECTIVE SAMPLING 
(Z) 
  
  
  
Terrain relief modelling is the process of T 
extraction and sampling of the spatial location of J 
points or lines on terrain surface. 
  
MAPPING Z-SET INTO GRID DOMAIN AND 
  
  
1. OPTIMUM SAMPLING PARTITIONING INTO PATCHES 
T 
For large scale (detailed) DTM, optimum sampling Y 
can be a combination of semi-automatic progressive 
sampling (PS) of more homogeneous (ll-set) supple- PROGRESSIVE SAMPLING 
mented vith manually controlled selective sampling (x) 
  
  
  
(SS) of the skeleton information (Z-set). T 
l 
  
CS combines selective sampling (I-set) with 
progressive sampling (m-set). The aim is to EVALUATION 
portray terrain relief faithfully without 
excessive redundancy of the sampled information. 
  
  
  
Fig. 1: Main stages of optimum sampling 
The input consists of photographs of the terrain 
and the corresponding control data. The specified 
  
  
DTM grid is partitioned into square patches which 1.1.1 Selective sampling of anomalous regions. 
act as the vorking units. General procedure for data preparation and feature 
; . : extraction for Selective Sampling of distinct 
The output of CS is an incomplete regular DTM grid morphometric features and "anomalous terrain 
vith density adapted to local terrain roughness, (E-information), for the purpose of subsequent PS, 
supplemented with the skeleton information is treated in (Makarovic, 1976) 
(E-set). The original selectively sampled Z-set is 
preserved in the data base. Because SS is subjective, it needs to be 
; systematised. To attain a balance between SS and 
To attain an optimal procedure the basic functions PS and a smooth operation, some rules have been 
should be logically structured and adapted into formulated. These represent the RULE BASE for SS 
the system component. and CS. 
The four main stages of optimum sampling are shown 1.1.2 Rules for selective sampling. From the 
in figure 1. results of the experimental tests applied to ideal 
geometric generated primitives, their composite 
1.1 Selective sampling surface and to real terrain relief, the following 
rules have been extracted for selective sampling 
General. (as integral part of composite sampling): 
SS is carried out manually to portray and /or 
isolate or exclude the anomalous regions in The general procedure for segmentation, extraction 
terrain. It is applied to abrupt changes in and selective sampling of the terrain relief 
terrain slope, peripheries of water surfaces, features, inside anomalous regions, are explained 
clouds and image areas, with a poor stereoscopic in (Charif, M. 1991). The corresponding rules for 
hold, etc. Basically, SS is a subjective method of selective sampling of the terrain relief features 
portraying the skeleton of terrain relief and of are as follows: 
isolating the anomalous regions. 
Features approximating the geometric primitives 
are sampled only up to the following limits: 
77