Hi, what do you thin is better? Ordinary camera* connected to arduino ( if that is not enough, Rasp Pi will be better? ) or kinect ?
I need that to recognize a object and sending video to the pc or mobile.
*camera like pixy or something like that :)
Depends on your needs
If you are just using the camera for the purpose of recognition and not any 3d manipulation, then Kinect is probably abit overkill.
I would suggest the two following approachs:
If you are going to be recognizing the objects by colour then pixy camera/arduino setup will be perfect, if you are planning on recognizing the object by other features, such as a logo, or face recognition, then I would suggest a raspberry pi and use the CSI camera that is designed for the pi.
The raspi/csi camera setup will be perfect for both of them, and will probably have more uses, but it depends on your budget and needs (not that they are very expensive, about £45 all in all)
With regards to actually sending the information back to some form of mobile device, I have no experience doing this with an arduino, although I suppose its possible, however raspberry pi will make the whole process easier. (You could just upload to a database or something)
Pi+camera is more adaptable.
Pi+camera is more adaptable. Kinect is actually a computer with an infrared camera and a normal camera, doing matrix computation with a know calibration and fixed focus point.If you need a 3d point cloud it may be the fastest way, but once you have the raw data you still need to find objects, you still need another computer to do it.
Raspberry and Pixy are stand alone, no 3d scanning, but fast enough to follow blobs of color or find geometric shapes. To recognize object from images look about artificial neural networks!
There are several
There are several interesting projects that are available to recognize objects. The OpenCog project uses DeStim for this and they have a version written in Python if you want to experiment with it. It doesn’t seem to be an ANN, but it sees an object and decides what it is by a voting mechanism.
It depends
A kinect is good for giving from .8 m to 4 m a 3d point cloud (ie x, y, z positions of any obstacles it senses). An Arduino lacks the memory to manage the sheer volume of points it would create so requires RasPi, PC or some such… You would need something to then interpolate what it is seeing to give you guidance. ROS (robot operating system at ros.org) can relatively easily be used to do this. With some work, it will do navigation for you as well. You tell it where you want it to go and ROS will pick the best path for you based on the 3d map. Pretty cool stuff.
A kinect also has a camera which you can use to identify your object vis OpenCV, and the 3d point cloud will identify its position accurately up until it is a few feet away. For myself,I would do the kinect for what you are talking since they are $25 at gamestop.com versus $70 for the Pixy. But it also depends on your skillset and what you are comfortable with. A pixy with an Arduino might be ideal for you since less to learn to use it.