252 
COMPOSITE SAMPLING 
USING A MINICOMPUTER SUPPORTED ANALOGUE INSTRUMENT 
Huurneman,G. 
Tempfli, K. 
ITC, The Netherlands 
Commission II (WG II/1) 
ABSTRACT 
A system for semi-automatic data collection for digital elevation 
models (DEMs) has been developed at ITC. An analogue stereoplotter 
(Planimat), equipped with rotary encoders and step motors, is inter- 
faced with a minicomputer (PDP 11/45). Distinct terrain features can 
be digitized in different modes. They supplement incomplete grid 
measurements. The local grid density is progressively adapted to the 
variations of terrain relief. This is achieved by near real-time 
analysis of previously measured elevations and synthesis, i.e., 
selection of further points to be measured. Positioning of the X and 
Y carriages of the analogue instrument is computer-controlled (opti- 
onally also approximate Z positioning). The paper describes the 
hard- and software components. Properties of the system are out- 
lined along with some proposals for improvement, extension and 
alternative system components. 
INTRODUCTION 
Composite sampling (CS) is a method of photogrammetric data collec- 
tion for DEMs, in particular for large scale applications [1],[4]. 
The objective of CS is to obtain,using a time efficient process, 
precise and comprehensive terrain relief data in a favourable struc- 
ture. CS unifies selective sampling (SS) and progressive sampling 
(PS). 
The basic principle of PS [1],[2] is iterative grid densification. 
Initial coarse sampling is followed by data analysis in near real- 
time, which permits synthesis for finer sampling.The result is a 
square grid with variable density,locally adapted to the fluctua- 
tions of terrain relief. Data are captured and stored per patch, a 
subunit of the stereomodel area which is specified prior to sampl- 
ing. Measuring is stationary; X,Y positioning is computer control- 
led, whereas height setting and triggering of recording is left to 
the operator. 
To increase the efficiency and reliability of PS, it is preceded by 
SS. Distinct surface features are sampled; these are breaklines and 
-points,and boundary lines of "anomalous regions" (clouds, water 
surfaces, dense forest, areas of poor stereoscopic hold, highly 
irregular and extremly regular surfaces). SS requires photo-inter- 
pretation. The data are acquired per model, in dynamic or stationary 
mode depending on the type of feature, and are stored in vector 
format. Following the integral approach of CS |3|, the selectively 
sampled terrain features are converted to raster format and partiti- 
oned to patches. They are utilized in PS for analysing local relief 
variation. 
253