in 
of 
or 
le 
le 
le 
commission errors (included in a class). Therefore, 
the misclassification matrix seems to be a better tool. 
Completeness will report on missing features due to 
inattention of the operator or obstructed objects. For 
photogrammetric control the superimposition 
technique is the most reliable. Completeness is 
checked based on specifications with additional rules 
such a minimum length, minimum width, minimum 
area, etc. Features or areas where field 
completeness is required can be given a special code 
during data collection, facilitating in this way the field 
operations. In addition, clear specifications are 
required for minimum completeness percentage, as 
well as producer and consumer risk. 
Logical consistency can only be checked to some 
extent with available CAD software. More powerful 
systems where data collection takes place within a 
GIS environment offer more possibilities for on-line 
checking. Temporal accuracy refers to the currency 
of data. This type of information may be critical for 
certain classes of features, subject to rapid changes 
and for applications requiring current data. 
2. QUALITY CONTROL AND 
PHOTOGRAMMETRIC FEATURE EXTRACTION 
Different strategies may be applied to ensure quality 
of a final product. The choice is between the final 
verification strategy at the end of a production line 
and the "zero defect strategy" where all the 
processes of a production line are controlled. This is 
obviously a better and more economic approach as 
it may lead to zero defect production; for this 
purpose, a quality system is required. It can be 
defined as a "documented organizational structure 
consisting of processes, procedures, resources and 
techniques with the objective to assure quality in that 
organization" (ISO). 
2.1 Quality control in a production line 
The "zero defect strategy" can be sudivided into three 
levels (Eslami, 1995): 
Level 1 is called the process control which 
continuously monitors the quality of the various 
aspects of data production and handling process. 
Decision rules need to be developed for each 
process based on quality parameters and quality 
standards. 
Level 2 consists of the data editing process in 
which every operation directly affects the quality 
of the data.lt is also a critical phase since many 
tasks are performed either in semi-automatic or 
automatic mode. 
Level 3 contains a monitoring system based on an 
acceptance sampling technique. Decision rules 
are required in order to accept or reject a work lot 
of output (e.g. maps) 
2.2 Process control procedure 
A control procedure starts by extracting values of 
relevant quality parameters which give an indication 
13 
of the integrity of performance in a particular process. 
For the orientation process well established accuracy 
standards are available for interior, relative and 
absolute orientation (e.g. maximum residuals errors 
in X, Y and Z). If a computed value is found to be 
acceptable, one proceeds to the next process, 
otherwise corrective actions need to be initiated 
(figure 1). 
  
  
Back to 
From vo process 
Extract a Corrective 
relevant action 
value 
  
  
  
  
  
  
  
  
Compare with 
a pre-defined 
tolerance value 
  
  
  
   
  
Acceptable? 
   
Next process 
Figure 1: Process control procedure 
A control chart is often used in production lines to 
detect deterioration of quality characteristics and to 
forestall failures of a process. It is a graphical display 
of a quality characteristic that has been measured 
and computed from a sample versus the sample 
number or time. 
As long as the graph is located between lower and 
upper control limits, the process is in control. A non- 
random pattern in this chart may be taken as an 
evidence of assignable cause (e.g. large mean Y- 
parallax on a particular instrument). 
2.3 Photogrammetric feature extraction: 
processes and products 
A photogrammetric production line can be subdivided 
into 6 to 8 processes, depending on the type of 
product to be delivered (e.g. spatial data, DTM, 
orthophoto). For each process and the derived 
product, a quality control procedure needs to be 
developed. 
The quality of a product can be described by a 
certain number of measurable parameters, whose 
values will compared to given standards 
(figure 3). 
2.4 Quality parameters and standards 
In the past, quality control strongly emphasized the 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996