Environmental Engineering Reference
In-Depth Information
(a)
(b)
(e)
(c)
(d)
FIGURE 3.8
The proposed approach applied to the VHR images over Boumerdes to detect building and damages. (a) A
sample of the original image through which the buildings are classified (b) and refined (c) by using morphological operators.
Adding shadow information helps reduce the buildings to the actual situation (d).
of this last method is presented, and a comparison with the GIS
manually extracted from the pre-event image is shown.
The results obtained show that all the buildings present in the
image have been identified, and the same can be said of the dam-
aged buildings, while only 3 out of 73 buildings have been mis-
classified as damaged, giving rise to an overall accuracy of 96%.
Using several examples, it has been shown that per-pixel
analysis is no more the right way to exploit the information in
the data when the spatial resolution is finer than the relevant
elements in a scene. Per-object or, more generally, spatially-aware
and context-sensitive approaches are required and should rely
not only on remotely sensed data, but on the
apriori
knowledge
about the test site and its environment.
As an example of this way forward, change detection in case of
severe earthquakes using mathematical morphology operators to
refine pre- and post-event maps and combine their results in an
effective way has been proposed. Results are encouraging from a
quantitative point of view. However, the main aim of this chapter
is to provide some ideas and a general framework, leaving
to the many researchers working in the active field of urban
remote sensing the freedom to develop their own specialized
procedures.
Conclusions
In this chapter we have discussed multiple limits and challenges
of VHR optical imagery for urban-related applications. We
also reviewed some algorithms suitable to reduce some of the
problems and cope with some of the challenges related to urban
mapping and change detection at the building level.
Search WWH ::
Custom Search