topic Re: detect dark spots using camera in Machine Vision

detect dark spots using camera

Arjun247 — Mon, 29 Mar 2010 15:28:49 GMT

I have taken a snapshot of a light source. I am posting a sample of

bad light source. It has dark spots in it. I just need the area of dark

spots of the circle. The processing I did to get an equalized image is

acquisiton-->extract luminance plane-->imaq mask and extract-->threshold-->imaq equalize.

i used vision assistant for the above steps.

The images are attached to this post. I did use count and find objects, but results are not

that accurate. Does anyone know a better way to get more accurate results. Just let me

know if anyone knows a better way of finding all the dark spots.

Re: detect dark spots using camera

Andy_C. — Tue, 30 Mar 2010 18:35:40 GMT

Hey Arjun,

How many spots were you able to locate? Could you post your script?

Re: detect dark spots using camera

muks — Wed, 31 Mar 2010 05:22:57 GMT

Is the blue patch that you are calling as a black spot? Will the perfect image have a perfect white circle? One method you ca use is detect circles in binary image and depending on the diameter you can say whether you have any black spot or not.

Re: detect dark spots using camera

Arjun247 — Wed, 31 Mar 2010 18:51:15 GMT

Muks

yes a perfect image will have perfect white circle. I am attaching the code as well as

original image and snap of my output. I got this code from vision assistant. In original

image the innermost circle is what I am masking and extracting.

When you will open my output image you will see that the dark spots near the border

of circle go unnoticed. The reason could be because of no border. I need to find the area

of all the dark spots, and thats it. If anyone feeds the my original to my code you will see

the result of what I see on my output. I appreciate your help. In circle the slight dark spots

are what I am lookingto detect.

I did try what Muks suggested os using circular edge detection, the result was interesting but

not what I need.

Re: detect dark spots using camera

muks — Thu, 01 Apr 2010 05:51:08 GMT

Before circular detaction you can try morpological operation namely dilation and increase the iteration till you get satisfactory result. Can you attach the vi for 8.0 or post a screen shot?

Re: detect dark spots using camera

Arjun247 — Thu, 01 Apr 2010 15:01:50 GMT

I am attaching the snapshot of the code. I am trying some experiments.

I will keep you posted.

Re: detect dark spots using camera

muks — Fri, 02 Apr 2010 05:46:17 GMT

Post a good image and a bad one so that we can also work on that.

Re: detect dark spots using camera

Arjun247 — Fri, 02 Apr 2010 13:33:50 GMT

I am attaching the good image to thia post. Here w go.

Re: detect dark spots using camera

muks — Fri, 02 Apr 2010 13:39:38 GMT

And again will it be ok if I give you a solution to differtiate the good and bad? Is some measurement involved?

Re: detect dark spots using camera

muks — Fri, 02 Apr 2010 14:05:25 GMT

Try this

Re: detect dark spots using camera

muks — Fri, 02 Apr 2010 14:18:55 GMT

Helped? Update as I am planning to leave now for the weekend........

Re: detect dark spots using camera

Arjun247 — Fri, 02 Apr 2010 18:06:04 GMT

well. What I was trying to say was the measurement should be based on area of dark spots

or black area. If black area is more than 10% of the area of circle then it should fail. Thats

what I am trying to do. thanks for your quick response.

Re: detect dark spots using camera

Arjun247 — Fri, 02 Apr 2010 18:52:01 GMT

You are filling up the holes in your code. That is what I dont want to do.

The area of black spots is the most important information I need to get.

Also the dark spots need not be in the edges of the circle. They could be

anywhere in the circle.

Re: detect dark spots using camera

muks — Mon, 05 Apr 2010 05:20:15 GMT

I had asked you this

And again will it be ok if I give you a solution to differtiate the good and bad? Is some measurement involved?

Arjun247 wrote:
You are filling up the holes in your code. That is what I dont want to do.
The area of black spots is the most important information I need to get.
Also the dark spots need not be in the edges of the circle. They could be
anywhere in the circle.

Again What i had suggested detects the blak spots for sure isnt it? Do you want to overlay them?

Re: detect dark spots using camera

Arjun247 — Mon, 05 Apr 2010 13:34:53 GMT

Yes I want to overlay them. I am sorry if I didnt mention it.

I wanted to overlay the dark spots and also calculate the area

and thier percentage in total fibers. That is what I am planning

to do.

Re: detect dark spots using camera

muks — Mon, 05 Apr 2010 13:40:43 GMT

This is entirely different to what we have been trying to do so far.....

Re: detect dark spots using camera

Arjun247 — Mon, 05 Apr 2010 15:55:54 GMT

Muks I am sorry if I confused you. If you read my first post it said clearly what I was trying to do. I needed percentage area of dark spots in the image. Arjun