Difference between revisions of "RCCamera"
(→Camera We Have) |
Cwehmeyer3 (talk | contribs) |
||
(8 intermediate revisions by 3 users not shown) | |||
Line 21: | Line 21: | ||
** [http://www.videredesign.com/docs/dcam-cam.pdf Manual] | ** [http://www.videredesign.com/docs/dcam-cam.pdf Manual] | ||
** [http://www.videredesign.com/docs/dcam-soft.pdf API Manual] | ** [http://www.videredesign.com/docs/dcam-soft.pdf API Manual] | ||
− | ** | + | ** [[Image:Dcam30a.zip|DCAM-L Software]] |
===Camera Resources=== | ===Camera Resources=== | ||
Line 29: | Line 29: | ||
* [http://sourceforge.net/projects/unicap Sourceforge Unicap] | * [http://sourceforge.net/projects/unicap Sourceforge Unicap] | ||
* [http://sourceforge.net/projects/iidcapi Sourceforge IEEE1394 Digital Camera API] | * [http://sourceforge.net/projects/iidcapi Sourceforge IEEE1394 Digital Camera API] | ||
+ | * [http://damien.douxchamps.net/ieee1394/libdc1394/v1.x/faq.php DMA info for reading from camera] | ||
+ | * [http://www.linux.com/howtos/libdc1394-HOWTO/concepts.shtml libdc1394 article] | ||
+ | * [http://mi.eng.cam.ac.uk/~er258/cvd/index.html A computer vision library] | ||
==Lens== | ==Lens== | ||
* 68 degree FOV (with one camera - not really practical) | * 68 degree FOV (with one camera - not really practical) | ||
* 35 degree FOV (with two cameras) | * 35 degree FOV (with two cameras) | ||
+ | |||
+ | [[Category: RoboCup]] |
Latest revision as of 20:52, 13 June 2018
Camera
So, after speaking with the folks at the US Open, I have some information to report about the cameras used. The CMU team used two of the smallest Sony prosumer three chip cameras with wide angle lenses. They did not stream from firewire, but rather used S-Video and a capture card per camera. They used the deinterlaced frames to get a apparent speed of 60 FPS. Consequently, they need the enhanced color reproduction of the three chip cameras because they are using the deinterlaced frames which implies they have about half the resoloution (320 x 240) of interlaced frames (640 x 480). That is, the ball makes up only about 12 pixels with the deinterlaced frames, and accurate color reproduction is a necessity to be able to pick the ball out of the background noise.
So, with a quick enough camera, the color reproduction should not matter as much.
For professional machine vision cameras, the following look pretty good.
Pro Cameras
- Basler
- A601fc (60 FPS) or the A602fc (100 FPS). They are VGA (640 x 480) color firewire cameras which are linux compatible.
- Point Gray has some promising cameras (and generally better pricing).
- Flea 60 FPS at VGA 12 bit color
- Flea 2 60 FPS at VGA 12 bit color
- Dragonfly 2 60 FPS at VGA 12 bit color onboard RGB, YUV etc conversion.
- Scorpion 60 FPS at VGA 12 bit color
Camera We Have
- Videre DCAM-L
- 4 mm CS mount lens
- 30fps color
- 1/4" Sony HAD CCD sensor 3 Bytes/Pixel
- Manual
- API Manual
- File:Dcam30a.zip
Camera Resources
- Linux Driver Page Really good!!
- CMU 1394 page
- Sourceforge 1394
- Sourceforge Unicap
- Sourceforge IEEE1394 Digital Camera API
- DMA info for reading from camera
- libdc1394 article
- A computer vision library
Lens
- 68 degree FOV (with one camera - not really practical)
- 35 degree FOV (with two cameras)