Model and resolution File size by mesh size
5 x 5 m 12.5 x 12.5 m 25 x 25 m
160161_15 µm
Cannot be calculated
2302 Kb 582 Kb
160161_30 µm 2984 Kb 686 Kb
160161_60 µm 2983 Kb 686 Kb
161162_15 µm 3257 Kb 821 Kb
161162_30 µm 3929 Kb 848 Kb
161162_60 µm 3929 Kb 848 Kb
Table 6.2: File sizes for models in scale 1:25,000.
It will be seen, from table 6.2, that the files which are expected to have the same size do not. For model 160161, the file sizes with equal mesh size should be the same, regardless of the resolution, and it ap-pears from the table that this is not the case. Likewise, it will be seen that the file sizes for model 161162 are not the same, regardless of the resolution.
Appendix G includes a table where all the file sizes are indicated, and it is seen here that files expected to have the same size, regardless of scale and resolution, do not fulfil expectations. The different file sizes may have several causes, for instance that the calculation has stopped before it is finished, unequal mesh sizes, or that the generation is done over unequal area sizes.
6.1.2 Control of the file sizes
GeoCAD has been chosen as the visualisation tool for the investigation of whether the individual files are fully calculated and have acquired the wished-for mesh size. (GeoCAD is a Danish developed CAD pro-gramme for geometric data). This has entailed that not all results can be shown, as GeoCAD cannot han-dle very large data quantities. Therefore, there are only 16 plots in Appendix H, as three calculations could not be done, and eight could not be visualised.
It will be seen from the file plots that all calculations are finished, and that the grids have the required mesh sizes, just as they are, apparently, calculated over the required area. As it can thus be demon-strated that the calculations have the expected quality, the conclusion must be that Match-T’s visualisa-tion tool is not a reliable representavisualisa-tion of the grid calculavisualisa-tion, and that it cannot be used other than as a consultative visual indication of how far the calculation has run. This function will therefore not be dis-cussed further in the project.
6 Pre-analysis of the generated data
Figure 6.3 shows that there are different forms of coding. If figure 6.3 is subjected to a visual overview, four characteristics are demonstrated:
1) Boundaries are found along the model limits and along the northern edge.
2) The boundaries between points which are estimated as being defective (dark grey) appear in squares with varying interval points, and the boundaries are linear.
3) Grid points with a double code (medium grey) appear in a random pattern.
4) Areas with grid points estimated as OK are uniformly light grey.
Re 1):
On the background of a visual description, marked elements appear in the form of borders. A closer in-vestigation of the borders between the models and along the northern edge has shown that elevation data in these borders has been assigned the code 3 (outside the area). A further examination showed that the keyed-in LL and UR delimitations have not been observed. This means that the assigned code is correct! But why Match-T has included the points, when these lie outside the limits of the desired area remains uncertain. This phenomenon is not a singular case, but is valid for all the 144 generated files, which are included in this project!
These extra points have also led to an overlap between the models which was not wanted. Therefore, it has been necessary to develop a small programme which cuts out this extra elevation data from the ar-eas again, see Appendix D. After the cut-out of extra points, the file sizes were reduced and rendered
uni-Model 160161
(a)
(b)
(c)
(d)
(e)
Model 161162
Boundaries
Figure 6.3: The result for 1:25,000, 15 µm and 12.5 x 12.5 m. Grid points are indicated by codes, light grey = the grid point is OK, medium grey = grid point with double code, dark grey = the grid point is estimated as defective. Six types are emphasised, boundaries and the areas (a) – (e).
Re 2):
Grid points which are estimated as being defective (dark grey) appear in squares with varying intervals.
Examples of this are shown for the areas (b) and (c). As regards area (b), every second point is estimated as being OK, for area (c), every fourth point is estimated as being OK. The possibility that errors in the data, caused by, for instance, a correlation error due to objects in the terrain, should appear in fixed squares is estimated as improbable. Objects in Denmark do not appear in fixed squares. Furthermore, there are grid points between the points which are defective, likewise in a fixed pattern which does not appear probable either.
In addition, there are areas where every other or every fourth point in a dark area has been assigned a double code. An example of this is seen in (a). Here, it is the code for the boundaries which ”shines”
through.
Re 3):
Grid points which have been assigned a double code (medium grey) appear in a random pattern. An ex-ample of this phenomenon is shown in (d). That grid points with a double code should appear in a random pattern seems probable, as this could, for instance, be a reproduction of objects in the terrain such as hedges, ditches etc.
Re 4):
In area (e), all the points have been estimated as being OK. These areas are as required.
The five different areas (a)-(e) are summed up in table 6.3.
Area Description Comments
(a) Area where both every second and every fourth point have either the code = OK or a double code.
Improbable (b) Square where every second grid point has the code =
de-fective and every second point the code = OK.
Improbable (c) Square where every fourth grid point has the code = OK,
the rest the code = defective.
Improbable (d) Area where grid points with a double code appears in a
random pattern.
Probable (e) Areas where all the points have the code = OK. Probable Table 6.3: Comments on the chosen areas.
6.2.1 Summation
In the chosen grid in scale 1:25,000, resolution 15 µm and mesh size 12.5, there are many and large ar-eas where the coding is improbable (a) – (c). If these codes indicate that there are errors in the grid points, it is grounds for great concern as regards the use of automatic generation of elevation data. Only the areas (d) and (e) show results which are probable, and are wanted by the user.
A look at all the results from Appendix H shows that the number of dark grey squares is diminished, when the mesh size is increased. As regards a few results with mesh size 25 x 25 m, there are no dark grey squares, while with mesh size 12.5 x 12.5 m, there are large areas with dark grey. This indicates that the determination of a grid point is best when a large mesh size is chosen, as there are several interest points to determine the individual grid point, which corresponds with the theory, cf. Chapter 2, section 2.4.3.2.
The next step in the preliminary investigation is to find out whether the assigned codes are actually able to indicate gross errors in the data material. Before this can be done, possible gross errors must be identi-fied and located. For this investigation, the frame of reference and the PIL programme are used. The PIL programme is self-developed specifically for this purpose, a detailed description of the PIL programme is found in Appendix C. By means of the frame of reference, possible gross errors can be identified and lo-cated, and thereby isolated.