What "MICROCONTROLLER" that is best for IMAGE PROCESSING?

Raspberry Pi with a PiCam might be adequate.

Two popular options are Raspberry Pi and Beaglebone Black.

rPi has a better community support than BBB, but BBB has more power than rPi for these kind of projects.

Raspberry Pi board is less costly than BBB. But you adding the cost of PiCam to it. On the otherhand, BBB is costly than rPi, but can work with cheap USB cams, so can still keep the total costs down.

You can get a one-on-one comparison here : http://makezine.com/magazine/how-to-choose-the-right-platform-raspberry-pi-or-beaglebone-black/

Thank you for the suggestion guys..

but I wanted to know if it is doable to use IMAGE PROCESSING for simple waste segregation?

That question you can probably ask in this community : https://groups.yahoo.com/neo/groups/OpenCV/info

First of all, to perform visual inspection on an object or material it cannot be obscured or buried in a pile of trash, it must be separated or on top in full view, which requires some separation method. Full optical inspection may require viewing from multiple angles which requires multiple cameras or movement of the object. This implies visual inspection already requires more than just a camera and software to have any chance of success. There are some materials and objects that can be determined to be biodegradable by visual inspection, but there are many where visual inspection is simply inconclusive. Optical discrimination rate may be improved by adding extreme magnification and spectral analysis, which would still fall under the image processing umbrella, but may complicate the apparatus and material handling systems. For those object and materials that can be visually distinguished, performing the discrimination in software can be a very complex task and some choices will fail because the software is not sophisticated enough. So for general biodegradable decisions, image processing will not be very accurate, and many other sensor systems would be needed, working in concert, to improve accuracy.

However, there are special cases where image processing could be used successfully, and you might try one of these for your project. For example: image processing could distinguish whether an item is a cardboard cup or a plastic lid or a plastic straw, assuming these are the only types of trash on the conveyor.

To get to a good decision, you need to define your environment and the capabilities you need plus some timing goals.

Issues such as time to respond, color, B&W, or spectral images.  Resolution, spectral purity, and a host of other potential issues.

You also need to establish lighting requirements and sensor sensitivity.

As I have stated before, these kinds of questions are really impossible to answer without the background data.

To do a component selection you need to:

A.  Define you requirements

B.  Identify key separation parameters.

C.  Identify you candidates.

D.  Evaluate each candidate against your requirements.

E.  Select the least expensive solution that meets your requirements.

Never, ever, just compare candidates against each other.  You can get what you think is the best, only to find out that it does not work in your environment.

DAB

missing a key component and im sure you all know this

Speed of analysis

If this is a conveyor then the system must analyse in almost real-time, yes there is a lag but it must be able to keep up with the conveyor, this will require a very fast processing sub system or many small systems working on concert

all the other points are perfectly valid also, we just need to add the dimension of performance, this in itself would probably eliminate things like PI, BBB etc. From a proof of concept they will work but for production it would need to be massively scaled up, or the conveyor slowed down

Peter Oakes wrote:

 

missing a key component and im sure you all know this

 

Speed of analysis

 

If this is a conveyor then the system must analyse in almost real-time, yes there is a lag but it must be able to keep up with the conveyor, this will require a very fast processing sub system or many small systems working on concert

 

all the other points are perfectly valid also, we just need to add the dimension of performance, this in itself would probably eliminate things like PI, BBB etc. From a proof of concept they will work but for production it would need to be massively scaled up, or the conveyor slowed down

Im not using a conveyor.  im going to make a two trashbins: 1 for the Biodegradable and 1 for the Recyclable.
on the top of the trashbin there is a CAMERA for image processing. If ever the person trying to put the "PLASTIC BOTTLE" to the biodegradable bin, the biodegradable bin won't open.
If the person put the plastic on the Recyclable bin the bin will open.

In that case the PI should be fast enough

God luck and keep us posted

ODROID XU4 or for really fast speeds get NVIDIA JETSON TK1