Image recognition robot; undergradute uni project

[sseb4]

hey im a student studying in Australia and need to construct a robot which views 1 of 4 symbols and performs a predetermined action. Eg. a cross (turn right),  a donut (turn left), triangle (stop), and a line (go straight).
We have been given an Altera Cyclone II FPGA, which we intend to store a soft core based processor using the NIOS II program(because it uses C++), but we can use verliog or Matlab simulink. Im not sure if anyone has heard of these or used these programming environments but this is what are predominantly taught at uni. The issue is we need a way of reading the images, we have come up with the two most obvious ways 
1. use an array of photo detectors or infrared detectors
2. use a real time camera and use pixel counting or matching methods
The information we have been given eludes to the fact that the symbol may not be aligned properly and because of the discrepancies of the detectors i am leaning towards the camera.
One camera i have found is
http://cgi.ebay.com.au/SMALLEST-SECURITY-COLOUR-SPY-CAMERA-TINY-PINHOLE-SIZE_W0QQitemZ330151368193QQihZ014QQcategoryZ20589QQrdZ1QQcmdZViewItem
Is there any camera or methods anyone can suggest? We have a budget of $115 AU excluding the FPGA, and need a few ideas on how to start this. We have never built a robot and don't have much experience, and need a frame or base to to actually hold it together, and Autocad and custom work is not much of an option. Oh yeah we have 6 AA batteries to power everthing
Thanks in advance
 

[Kohanbash]

Hi
Since you don't have a lot of robot experience I would recommend using a mobility platform either from a kit or based from the admins $50 robot tutorial.
If your allowed and if the background is clear of obstacles I would look online to find edge/line detection and circle detection code. Depending on the efficiency you can either run them on one core, or since FPGA is hardware based you can run multiple cores simultaneously. An example could be one core for driving, another for edge detection, and another for circle detection (if the background is super clean you can assume that if you see something and you don't pick up lines than its a circle).
Also you can run a core for mobility and use simulink (I have never used simulink on an FPGA) for the edge/line/circle detection since code is readily available on matlab's home page.

[sseb4]

thanks sdk32285 in reply to your question there are no obstacles, an i'll have a look at the $50 robot. After some research and input from some people who have PASSED the subject they all seem to say that image processing via a camera is over kill and far too messy, also considering that no one completing this project has been taught about image processing to begin with. So i'll opt for the array of photodetectors. If anyone knows of some fairly sweet photo - detectors/transistors/resistors of any sort that would work happily with a FPGA and 6 AA batteries please let me know where to find them cause i dont really know what is out there and i need to rely on some ppl's experience. They just have to be able to tell the difference between black and white in relation to what is directly in front of them.
Thanks again

[S. Karim]

just buy the pob-bot...