Difference between revisions of "2013-04-24 49 12"

From Geohashing
imported>Dawidi
(writeup and very weird photos)
 
imported>Dawidi
m (bullet points)
Line 14: Line 14:
  
 
The video was then reduced to a single column of pixels for each frame, and all those columns concatenated side-by-side, like so:
 
The video was then reduced to a single column of pixels for each frame, and all those columns concatenated side-by-side, like so:
Video to four-pixel-wide strips:
+
* Video to four-pixel-wide strips:
 
  ffmpeg -ss 00:02:10 -i "GOPR2687.MP4" -an -vf "crop=w=1080:h=4:x=0:y=360" -f image2 "part1\strip%08d.png"
 
  ffmpeg -ss 00:02:10 -i "GOPR2687.MP4" -an -vf "crop=w=1080:h=4:x=0:y=360" -f image2 "part1\strip%08d.png"
Rotate strips to vertical columns, and reduce to one pixel width:
+
* Rotate strips to vertical columns, and reduce to one pixel width:
 
  for %s in (part1\strip*.png) do @imagemagick-convert -rotate 90 -crop 1x1080+0+0 "%s" "part1s\line%s"
 
  for %s in (part1\strip*.png) do @imagemagick-convert -rotate 90 -crop 1x1080+0+0 "%s" "part1s\line%s"
Concatenate all columns :
+
* Concatenate all columns :
 
  imagemagick-montage -mode Concatenate -tile x1 "part1s\strip*.png" part1assembled.png
 
  imagemagick-montage -mode Concatenate -tile x1 "part1s\strip*.png" part1assembled.png
 
(With thanks to my coworker, whose 16 gigabytes of RAM proved necessary to assemble the second part in one go without massaging the swap file for hours - a custom program would've been much more efficient there than imagemagick...)
 
(With thanks to my coworker, whose 16 gigabytes of RAM proved necessary to assemble the second part in one go without massaging the swap file for hours - a custom program would've been much more efficient there than imagemagick...)

Revision as of 15:54, 24 April 2013

Wed 24 Apr 2013 in Regensburg, Germany:
49.0198471, 12.0619058
geohashing.info google osm bing/os kml crox

In a courtyard-like space between residential buildings on Hermann-Köhl-Straße.

dawidi

... went there in the morning around 10:30, on his way to work. As expected from aerial imagery, the courtyard was accessible from the east via a ramp and a passage; the hashpoint itself was pretty much exactly in the middle, behind a bench among a loose group of trees.

Since the hash was too easy - I cycle through that road almost every day - I decided to at least give you an unusual picture documenting the way there: a non-stop "scan-o-rama" stretching all the way from my home to the hashpoint. The camera (my trusty GoPro) was mounted in a cardboard box on my luggage rack, looking out to the left (and oriented in portrait mode to increase vertical resolution), and recording video at 60fps, while I tried to cycle more slowly than usual.

The video was then reduced to a single column of pixels for each frame, and all those columns concatenated side-by-side, like so:

  • Video to four-pixel-wide strips:
ffmpeg -ss 00:02:10 -i "GOPR2687.MP4" -an -vf "crop=w=1080:h=4:x=0:y=360" -f image2 "part1\strip%08d.png"
  • Rotate strips to vertical columns, and reduce to one pixel width:
for %s in (part1\strip*.png) do @imagemagick-convert -rotate 90 -crop 1x1080+0+0 "%s" "part1s\line%s"
  • Concatenate all columns :
imagemagick-montage -mode Concatenate -tile x1 "part1s\strip*.png" part1assembled.png

(With thanks to my coworker, whose 16 gigabytes of RAM proved necessary to assemble the second part in one go without massaging the swap file for hours - a custom program would've been much more efficient there than imagemagick...)

Effectively, the resulting image has the GoPro's usual stereographic (fisheye-like) projection on the vertical axis, but is projected orthographically (if I cycle in a straight line) or cylindrically (in turns)... with all kinds of wobbles and other artifacts, because I can't really keep my bike vertical and moving at a constant speed all the time.

Anyway, here goes:

Scanorama from a bakery near the hashpoint to the town center
The full image is 50000x720 pixels and about 13 MB
Scanorama from the town center to dawidi's office
The full image is 50000x720 pixels and about 12 MB


And as far as "regular" photos are concerned: