which board is appropriate for image processing?

[hamidkavianathar]

hey guys
I want to do a real time image processing on one of these (cubieboard, raspery pi or beaglebone, ..) board. I want to connect four cameras to it to monitor around. I want to install ubuntu and openCV on it. but I don't know which board is the best choice. it must support at lease for cameras and a hdmi (or vga) port.
could you please tell me what I should do?
thank you.
(I apologize for my bad English)

 

[123jack]

Unless you are doing something very simple indeed - none of these are suitable for image processing.
Perhaps if you explained what you want to achieve it would help to choose what you need.
("Image processing" could mean just about anything)

 

[hamidkavianathar]

Unless you are doing something very simple indeed - none of these are suitable for image processing.
Perhaps if you explained what you want to achieve it would help to choose what you need.
("Image processing" could mean just about anything)

thanks for the comment. I want to install four cameras around a car and a lcd in the car and some ultra sound sensor next to the cameras. then I want to build a smart car. I'll use sift or orb feature detectors ,... .

 

[andre_luis]

First of all, although having pretty powerful cores, none of these hobby-like boards you mentioned above are suited to perform real time image processing, particularly with complex algorithms like SIFT. For dealing with 4 cameras you should consider using some video encoder commonly used for surveilance to make the compression/store of the video at a HDD for further processing, otherwise you´ll have to seriously think about purchase an FPGA board.

 

[hamidkavianathar]

First of all, although having pretty powerful cores, none of these hobby-like boards you mentioned above are suited to perform real time image processing, particularly with complex algorithms like SIFT. For dealing with 4 cameras you should consider using some video encoder commonly used for surveilance to make the compression/store of the video at a HDD for further processing, otherwise you´ll have to seriously think about purchase an FPGA board.

thanks for the comment. but how can I use openCV instruction on a FPGA board? I'm familiar with FPGA and verilog but I don't know how I can run openCV and python instruction on a FPGA board.

 

[andre_luis]

If the performance is not an issue, and if the use of OpenCV library is a requisite of your project; yes, you can use some of these boards to do the job, but alternately taking pictures of each one of the four cameras and processing them in sequence, instead of processing the full video stream in realtime.

One easy option - considering the relative proximity of the cameras to the processing unity - could be using 4 USB Webcams with one of these boards. I particularly recommend the BeagleBone Black, due to the excellent didactic material available on the Youtube chanel, provided by prof. Derek Molloy.

 

[hamidkavianathar]

the performance is very import for this project. It must be real-time. the using openCV is not a requisite of my project. but openCV is a versatile library for image processing project.
I've decided to do it with an FPGA board because of its performance. it's also very cool for me.
could you please tell me how should I do it?
thanks.

 

[andre_luis]

According to the algorithms you have mentioned above, you certainly want to perform tracking patterns invariantly on scale and position, which is not a trivial task. The fact that you specified it have to do this job concurrently with 4 cameras, it is and additional overload that impacts on system requirements. I'm not familiar with FPGA designs, but I believe that skilled members will recommend you to first make a draft you what you intend, informing the key features, such as frame rate, encoding type, image resolution, etc...