what are or is the most effective real time image segmentation techniques & algorithms when dealing with a metallic cylindrical part being turned on a lathe machine?

real time image segmentation techniques

Answer 1

[Image Analyst]

If it's a computer controlled lathe, you know the progress and any time because you know the total time it takes to cut it. If you wanted to make another system to give you basically the same information you'd have to take like 100 shots, and then when you run it again, you can compare your image, through maximum absolute deviation or something like that to all 100 images and find out the percentage done you are. Like if the MAD was closest to your reference image #39 then you'd be 39% of the way done. Of course you know that also just simply from the time the picture was taken, the time you started, and the known completion time.

Answer 2

[Image Analyst]

How about thresholding? By the way, you forgot to attach your image like most people who want image processing advice. Did you seem image segmentation tutorial? http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

[Doctorknowledge]

You may check out the specific image l have attached, thanx

[Image Analyst]

Isn't this a computer controlled lathe so you know what you'll get every time? Also, that's not a good photo - it's at an angle and the cutter is in the way.

[Doctorknowledge]

yes it is a CNC lathe,also its correct you will be knowing the end product (stepped cylinder). But its just that l'm trying to to develop a vision based system using matlab, that will only estimate the progress based on the real time image that will be captured by a digital camera, My scope though is just to train the system so that when a camera is connected, the system will have learnt that is there be a match, it simply displays the corresponding % completion say 45%. Is this feasible on matlab?, if feasible, does that mean l'm supposed to take photos for training at an angle that doesn't show the cutter?