Vision sensor

Hi everyone! I am really interested in the implementation of object detection with my Arduino Mega. The question that comes to me is which vision sensor use. Since Arduino does not have the capability of processing the million of pixels that a vision sensor can have, I wanna ask you guys if you know a camera with which Arduino can handle its information incoming without overwhelming or a way to create an object detection platform with Arduino I found the Pixy camera really interesting, but I am listening more possibilites.

Thank in advance and greetings from Germany!

Roberto

Please don’t cross post, putting the same question in multiple places.
The duplicate has been deleted.

Also look at this as it also mentions vision stuff.

Thanks Riva, and sorry to the admin btw, I am newbie :slightly_frowning_face: :slightly_frowning_face:

It seems good but I dont know if with one shield Arduino can manage the dozens of megabytes that can come from a object recognition camera. Besides: "It's important to note that this shield will not work on the Arduino Mega", so it wouldnt fit my requirementes :'(

More suggestions?

To cope with the amount of data the shield will have a severely reduced resolution & colour pallet. It will work on a specific Mega clone but if you already have a normal Mega that's not much help. It depends on what you want to detect and with what degree of accuracy as to it's suitability. Another option would be Raspberry Pi

Riva: It will work on a specific Mega clone but if you already have a normal Mega that's not much help.

Specifically the one from SeeedStudio, but you have to (I believe) re-wire the board slightly and modify the code for the new pin(s) involved; basically the issue is that the "official" Arduino Mega doesn't bring out every pin on the ATMega2560 - and one or two of them are absolutely needed for the VE shield to work properly. Since the VE wasn't designed with the SeeedStudio Mega in mind, alterations to hardware and code are very likely necessary. I know of at least one guy on this forum who has done it, so if you search around you might be able to contact him about the mod.

Another potential (but much more difficult to implement) possibility would be to re-implement (and/or re-invent) the AVRCam (which was similar to the CMUCam line - also another possibility, I suppose):

http://www.jrobot.net/Projects/AVRcam.html

Here's the info to start with on the CMUCam:

http://www.cmucam.org/

Thanks guys, I really appreciate the help. The point is that my project requires the usage of the Mega2560 and a camera, so Raspberry Pi, Beaglebone or so are not possible.

Apart of the CMUCam, I had already looked throuhg this AVRcam and Blackfin, but they look like a little bit old, are they worth it?

Since I have good knowledge in Matlab, I wanted to use it for image processing, but maybe with the amount of data that can be transmitted to Arduino is not worth the trouble of using Matlab. In other words, could I change the algorithm and firmware of these cameras (for example for shape recognition) and still being "low-powerful" enough for communicating with the board?