[HOT] Haiti 2010

[Fred]

In reply to sam larsen mail and from my experiences at Unosat (rapid 
mapping) and Keyobs (project development, mining extract building with 
satellite imagery, Kibera, Somalia).

As part of the emergency:


We must reduce the processing time or find other partners who have 
capabilities to provide images correctly georeferenced.

For now, there are very few suppliers provide use (except in exeptionnal 
case such as in Haiti).
So in general we do what we can to georeference images rapidly to 
realize extractions (of floods, destruction, road...)
So it means that we have access to the imagery and got some specialist 
to work with.

So the ideal is to have strong relationship with the supplier or other team.


After manual extraction or semi automatic topics:

To extract manually information of roads, settlements, damage building, 
IDP is still a valid method .
For several reasons:
1 - Semi-automatic extraction ask a strong expertise and need to be 
cleaned.

In the final geometries of heterogeneous objects which remains difficult 
to read the information or calculated surface or accuracy problems 
centroid (as in houses, we need to catch the centroid).
2 - Then there is a question of time, we should be releasing information 
quickly. So the more you have information on the area and better 
priorities will be set up.


In practice, remote sensing software need to be improve to extract 
information correctly from imagery for humanitarian field of operation . 
As all the software and Remote Sensing algorithm have been designed for 
developing countries.

Only recently algorithm has been refined and look accelerating to deal 
with automatic extraction.

However it is clear that we should use the fastest techniques and 
adapted them to the field.
For humanitarian, manual digitizing is still important to respond 
quickly to a need.

And so we use the automatic extraction to retrieve information from the 
image and then you end by hand.

Example Haiti 2010: 200 000 digitalization point damage building. Unosat 
was in partnership with SWISS TOPO (National Geo...). And we did not use 
Remote Sensing for the extraction (very bad result and no time to 
improve it).
And impossible to recover the roads in this complex area urban.

Gaza: Land cover and damage building, semi extraction( software 
E-cognition). It is true that we could recover the roads, but it was 
quicker to do it by hand.

Sri Lanka: IDP movement, 65% semi-automatic method (pretty clean) and 
the rest by hand.

Kibera: Semi Automatic method and we finished by hand.
Besides, as many cities in Africa are the pixel values are mixed. There 
is a lot too clean.

OSM would have to launch a cleaning roads with their volunteers if they 
are to distinguish houses to the road J

At the end, it's very frustrating to have student years to finally 
extract the information by hand.

But when it's information are crucial to crisis management we should not 
hesitate a moment.

For the case of OSM, as you are dependent of imagery (so the body of 
crisis management). We must join with other partners who have the means.

And give the soul and symbol of social mapping.

For the crisis mapping seminar :
After having participated in Haiti alongside with HOT, these activities 
has given a windows to the people to be involve in this crisis and is it 
the most important.
If you look at the humanitarian action today, the population is no 
longer involved. Neither in short-term and a little in a long term 
(proposal phase, meeting phase, etc...).
The question is how we let people speak and not only the consultant, the 
UN agency for their country.

A + fred all the best from Pakistan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/hot/attachments/20100911/5d9c3911/attachment-0001.html>