to the matching process. Image data are distorted geometrically and
radiometrically (intensity). Geometric distortions originate in
internal imaging, external (camera) attitude, and in terrain relief
(and obstacles on it). Radiometric disturbances emerge from the
varying reflectance of the same objects at imaging, shadows, lack of
detail, periodic patterns, etc.
When matching, disturbances occur if the strategy, algorithms (and
/or their parameters) or the techniques are deficient.
II.4 Post-processing
Post-processing implies parallax editing and DTM calculations [4];
thus the influencing factors refer to these two sub-stages. Parallax
editing involves the following operations: inclusion of external
imformation, collective processing, and a-posteriori exclusion of
poorly matched regions (if not rejected before). They contribute to
improved quality of the results.
Several influencing factors are involved in supplementary DTM calcu-
lations. The corresponding operations are interpolation, resampling
(DTM), and compression. The corresponding loss in accuracy is usual-
ly balanced by gain in data utility.
Disturbances occur in both post-processing sub-stages. Prior to
parallax editing representative points of the matched pairs of seg-
ments must be defined. The transfer of such a point from one segment
into its conjugate implies a potential disturbance. Other disturban-
ces occur in absolute orientation, approximate algorithms for the
DTM calculations, filtering (for smoothing), compression, resamp-
ling, etc.
ITI. INDICATORS AND CONCEPTS FOR QUALITY ASSESSMENT
Quality control should be applied in all stages of the process. To
this end, appropriate concepts and theoretical tools need be selec-
ted and/or devised. These apply to the data inputs and outputs of
the sequential process stages (figure 1).
Distortions in data are caused by disturbances in the process. A
differentiation can be made between the pictorial and geometric
quality components of the data in the successive process stages
(table below). At image matching, the pictorial and geometric compo-
nents are merged.
Sequential Input | Digital Upgraded| Image Parallax | DTM
data image | raw data data match data data
Component
Pictorial x -X ed
Table: Components of quality for the sequential data sets.
Parallaxes and DTM data, however, contain merely geometric data.
Hence, the performance considerations should be structured accor-
dingly. In the following, an overview is given of the indicators,
concepts and tools for assessment of the performance.
346