tre
ler
he
ell
le
of
-
e
Darvishzadeh Varchehi, Roshanak
[73 Unchanged
== Changed
x
i
i6]
fs in the area
mop MIN À
Figure 2. The changed and unchanged 00
2.2 Digital Change Detection
In this approach, after scanning of the new photo, the same procedure for making the geo-corrected image as in section
3.2 was followed. In the later stage, after rectified photo registration, classification and segmentation were used to
extract roof changes, using existing knowledge contained in the GIS database. The final result (detected change) will be
compared with the result of on-screen digitized data. Figure 3 illustrates this approach.
Cal. of
roof area
Products / Input
Figure 3. An overview of the analytical approach
2.2.1 Scanning Photo & Georeferencing. A similar procedure was used as that explained in sections 3.2.1and 3.2.2
with the exception that the aerial photograph was
scanned and processed multi-spectrally. A resolution of 150 dpi was chosen to have the same accuracy in the source
data as in section 3.2.
2.2.2 Test Area and Data. For the next steps in this approach, three subsets were selected from the georeferenced
images. The subset images consist of 520 lines and 500 columns, almost covering the middle part of the aerial
photograph (an area of 41600 m?). Classification and segmentation was done only for this part. Figure 4 shows the
colour composite of the subsets for the test area.
From the updated (1994) layer of the digital data (updated roofs of buildings) that was detected in section 3.2, the test
area was selected (reference map). The boundary of one of the subsets was identified and only the roofs, which would
fall, in that area were selected. This data will serve two purposes:
1. Knowledge acquisition from existing GIS (e.g., roads) for the knowledge based classification (improvement)
2. Evaluation of the classification and segmentation results for extraction of roofs.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 315
