Difference between revisions of "OSMCAM"
(→Modifications for the Elphel community) |
|||
| (One intermediate revision by the same user not shown) | |||
| Line 8: | Line 8: | ||
A few weeks we are already mentioning of collecting all photos processed in JOSM on http://www.openstreetphoto.org, if a camera was able to know what frames are interesting and stores them that would be ''nice'', if we can progress with it and actually do ''detection'' and ''classification'' on the camera itself that could be a reason to build other projects from it. | A few weeks we are already mentioning of collecting all photos processed in JOSM on http://www.openstreetphoto.org, if a camera was able to know what frames are interesting and stores them that would be ''nice'', if we can progress with it and actually do ''detection'' and ''classification'' on the camera itself that could be a reason to build other projects from it. | ||
| + | |||
| + | Creating an embedded ''editor'' or ''viewer'' on the camera that allows a user to see the collected GPS traces and is able to walk over a trace showing different photos. --[[User:Skinkie|Skinkie]] 09:03, 23 August 2008 (CDT) | ||
== Modifications for the Elphel community == | == Modifications for the Elphel community == | ||
| Line 24: | Line 26: | ||
Feature recognition in Dutch traffic; On going project, probably takes more than 3 months to get it to work. | Feature recognition in Dutch traffic; On going project, probably takes more than 3 months to get it to work. | ||
| + | |||
| + | == Current Work == | ||
| + | We are implementing a site for offline recognition here are our initial results: http://sift.openstreetphoto.org/matches/ | ||
Latest revision as of 03:51, 26 August 2008
Contents
Who are we
OSM NL is the Dutch chapter of OpenStreetMap. We are resource rich, with gifted developers, cartographers, writers, geo-engineers and inspirational speakers. We have gone were no country has gone before, a near complete map. But since the world changes and we still would like to innovate, some ideas were brought up. On of them is modifying an existing camera such as the Canon with the CHDK firmware. We have had some success with this, but the basic problem with still camera's is; the tend to overheat easy, leaving all photos very white. Processing sadly, isn't possible, but it is easy to put the cam in 'auto-shot' mode and shapshot a complete trip.
What do we want to do for ourselves
The idea of using a camera for geotracing is not new, it is relatively easy to link photos to a GPS position (just by cross-referencing the time in our geo-editor JOSM but much more interesting would be if the camera could actually point in the direction a photo was taken using a compas + gps.
Automating the process of tagging is another interesting idea that was prototyped before. Technically it is doable using OpenCV to scan for street signs and street names. The most easy solution would be a camera on a helmet or some more advanced to put a parabolic mirror in front of the camera so it will make 360 degrees shots. This could be an interesting setup, and it might be easy enough to use on a bicycle. This all can be done with ordinary off the shell components but the interesting thing of the Elphel would be that it is integrated.
A few weeks we are already mentioning of collecting all photos processed in JOSM on http://www.openstreetphoto.org, if a camera was able to know what frames are interesting and stores them that would be nice, if we can progress with it and actually do detection and classification on the camera itself that could be a reason to build other projects from it.
Creating an embedded editor or viewer on the camera that allows a user to see the collected GPS traces and is able to walk over a trace showing different photos. --Skinkie 09:03, 23 August 2008 (CDT)
Modifications for the Elphel community
Since we have a great deal of experience with GPS devices and (the frustration of) communication with them we could integrate off the shelf GPS'es as plug and play for photo and video annotation.
We would like to port the Cherokee Webserver [1] to the Elphel, allowing it to function as application platform for camera control and image transfer.
Convert scripts to bare C programs.
...and just for the sake of 'because' I can, I want to make a custom Linux image ;)
Timeline
GPS test making them plug and play; 2 weeks of development time to probe each serial over usb port to be seen as GPS device.
Cherokee; 1 month of development time to make extensions and to convert existing software to handlers.
Feature recognition in Dutch traffic; On going project, probably takes more than 3 months to get it to work.
Current Work
We are implementing a site for offline recognition here are our initial results: http://sift.openstreetphoto.org/matches/
