|
Configuration and cameras for roborealm
Annamaria Rossi [2 posts] |
13 year
|
HI i want to use roborealm to track objects , make obstacle avoidance, recognize shapes and numbers, count objects and manipulate them using robotic arm with maximum speed and fluidity.
i have planned to use 2 sony snc ch-120 cameras http://www.sony.it/biz/product/nvmfixedcameras/snc-ch120/overview
or two vivotek cameras http://www.vivotek.com/products/model.php?network_camera=ip7151
attached to a ethernet switch or directly into a computer.
so my questions are these...
can i use two camera at the same time in stereo vision with RR?
what kind of configuration i need in order to make all these things ?
initially i wondered about core i7 960 configuration with 16 gb of ram
but i'm not sure if it is ok or too much for roborealm
|
|
|
|
Anonymous |
13 year
|
Zippo,
Nice cameras! Those should produce some very nice images.
I would double check before purchasing that they can be interfaced in some way to RR. If they are IP based you may want to find a demo online of the cameras and check that you can get the needed URL and feed that into the HTTP_READ module to ensure that the images can be used in RR. Alternatively, if the vendors have a directX driver for those cameras you should be ok using that (this would allow any windows application to access those cameras).
You can use two cameras at the same time to create a stereo pair in RoboRealm. This can be done using two HTTP_READ modules OR using the Marker module you can switch between directX camera drivers. If you look at the Stereo module you will see that it accepts directX cameras or images saved to memory using the Marker module which can come from other sources like the HTTP_READ module.
Using an i7 would be fine ... the more CPU power you can buy the better. In terms of RAM you would need enough for about 5x the largest image that you will capture from the camera. In most cases 16GB is way more than you need. Most of the times RR works fine with 4GB but can also go down to 128Megs if your image is smaller. It will also depend on what you are trying to accomplish.
What image size will you be working with?
First step is to get those camera's connected to RR. It looks like both support MJPEG but you will need to find the right URL to ensure that the HTTP_READ module works. If you can find a demo site where these cameras are used we can help you find the right url.
STeven.
|
|
|
|
Annamaria Rossi [2 posts] |
13 year
|
thank you for reply
i would set up roborealm in stereo vision using 1024x768 images from cameras at 30 fps.
i will use 8 gb of ram on the computer.
my goal are these:
1 follow coloured objectcs
2 make obstacle avoidance in a generic place
3 make obstacle avoidance in a generic place using code markers.
4 count small and medium objects recognizing their shapes
5 manipulate objects using two dynamixel based robotic arms
do you think i need core i7 960 in order to make these things working fluidly or i can try with core i5 650?
what is the better camera for this vision applications between sony snc-120 and vivotek cameras?
do you know any other ccd cameras that can be interfaced over ip with higher image quality; less noise and 30 fps speed ?
|
|
|
|
Anonymous |
13 year
|
Zippo,
You can accomplish those tasks using a i5 but it is a question of how quickly you'll be able to process frames. While the camera may produce 30fps doing stereo on a regular pc at that rate is probably not possible. To keep those high frame rates you will probably need to work with a GPU like NVidia. Stereo is NOT a light analysis to perform and will reduce even the most powerful PC to a slow frame rate depending on the size of the image being processed.
So you should pay for as much CPU power as is within budget and then play with the image size (i.e. smaller) if you need the higher frame rates. Note that the tasks above do not appear to require a 1024x768 image. For most tasks 320x240 seems to work well and keep the CPU speed up.
We have not worked with either the Sony or the Vivotek camera and thus can't really offer much in terms of recommendations. This goes for any other high end ccd cameras. We typically focus on lower end cameras as that provides us with a more robust software testing platform. If the module works with a lower quality camera it almost always improves with higher quality ones. Again, in my opinion your tasks do not appear to require very high quality camera. If you have doubts in ordering the camera get the machine first and a $50 webcam and see how well that works. That may provide you with insight as to what you need to purchase in high quality cameras.
STeven.
|
|
This Thread