I guess you'd need to be more specific about your intended application. For example, meat industry is using optical methods to measure a color of meat that can be related to certain biochemical and metabolic processes. Then, it can be linked to very specific parameters established for meat quality. You can also detect water content etc. What exactly do you want to detect, what segment of food industry are you targeting and how do you define the term "quality" for these foods?

I am fully agreed with Mr Serge Grabtchak that it depends upon the type of food because Optical methods are related to the colour variations in food which are transformed in specific scales .. food type introduces different parameters which drives the colours.

I have a relative who is a mechanical engineering student at MIT. Last summer he worked in the MIT Media Lab where his task was to develop a system that could remotely monitor the location of a dog in a room. I think his work could be modified to do what you need, and it would relatively inexpensive to develop.

Here is what he had to say:

"For my project, I was using a library for Python called OpenCV. I programmed in Python because someone else started the project in Python, but if you wanted to use the library I would recommend programming in C++ since there is more documentation for the library. It's an open source computer vision library that's capable of doing all sorts of things from object recognition and 3D mapping to image processing. I was using the image processing part of it to threshold each frame from a stationary camera with a known background to create a contour around the parts that changed in order to find the location of the dog. I used a standard webcam with a wide angle lens so that I had as wide a viewing angle as possible, but OpenCV can work with all sorts of different input devices, I think. It's a pretty powerful tool. As far as wireless, I'll bet that could be done with Bluetooth or a wireless card on an Arduino or something like that.

The main trick would just be figuring out how you could distinguish the information you're interested in. In my case, I took some frames of video and played with them in Photoshop to figure out what processes I would have to use in order to create a mask over the dog.

Sounds like you could certainly use OpenCV for that problem."

I hope this is helpful.

I will find many interesting devices with Arduino (www.sparkfun.com).