LabVIEW

cancel
Showing results for 
Search instead for 
Did you mean: 

Detecting pills

Hi

I want to detect broken pills, How can i do that ?Bad Image 2.png

0 Kudos
Message 1 of 3
(2,203 Views)

Dear Samihamid,

 

your question is very brief, so is my answer. Vision Development Module will do, then use one of examples for this toolkit which deals with image recognition. Such program is quite simple when using VDM.

0 Kudos
Message 2 of 3
(2,189 Views)

Indeed, not much to go on... It would work better if you make an attempt, and ask for help on the details.

 

It highly depends on details. And with machine vision, there are a lot of details:

+ Are the pills always on the same locations?

+ Are the pills always the same color?

+ Are the pills always the same size?

+ Are the pills always the same shape?

+ Is the lightning constant (enough)?

+ Can the background color be changed?

+ Can the lightning be changed? (It's OK, but isn't great.)

+ Can the camera be focused? (The text seems blurry)

 

There are so many ways to get to something that works, and only the correct knowledge of the details will give the 'best' solution.

 

What is the best solution?

+ Is there a time restrain?

+ A CPU or memory limit (of course there is, but is it expected to be a bottleneck)? 

+ Restrain on software to use?

 

What is the expected result? 100% correctness is not likely to happen. You should specify the acceptable number of false negatives and false positives. Usually only one of them can be 100%, at the expense of the other value...

 

Vision isn't easy. First step is to optimize the acquired image. For instance, filtering out the red light will make (only) the pills black. That could greatly simplify the thresholding. With a color camera that is less significant though. A circular polaroid filter could reduce the reflection of the metal.

0 Kudos
Message 3 of 3
(2,164 Views)