How to remove the pole from the NIR image

How to remove the pole from the NIR image

Answer 1

[Image Analyst]

Since it's in a known location, simple create a binary image as a template/mask and mask it out

irImage(mask) = 0;

[Malini]

Thanks image analyst. Is there any way to extract the plant by using only one mask as you mentioned above to remove the background, pole and pot in one go?

[Image Analyst]

Assuming the camera position is fixed, the pole will remain fixed and thus you can simply mask it out. However if the plant goes in front of it, you'll have to not mask it out. You can't mask out the plant and pole in one mask since they seem to be about the same gray level.

[Malini]

irImage(mask) = 0;

In this step i am losing some of the plant portion. Is there any method like texture based segmentation to extract only the green plants?

[Malini]

Also in some images, the pole is not completely removed. some of the plant pixels are matched with the pole pixels

[Image Analyst]

Yep. Just what I said. The best option is to paint the pole white or change the material so it doesn't reflect the same amount of energy as the plants. Or perhaps you could combine it with an RGB image where you could somehow use color segmentation to get non-green things.

[Image Analyst]

Malini's "Answer" moved from being an Answer to here since it's really a comment to me rather than an answer to her original question at the top:

"From this gray image,how to segment non green pixels..."

[Image Analyst]

You don't. You need to also capture an RGB image from the same viewpoint. So two images from cameras mounted side by side. Or, cheaper, simply paint the pole white or use a pole of a different material.

[Malini]

I already have an RGB image. Will try to change the pole color

Answer 2

[Malini]

From this gray image,how to segment non green pixels...

[Image Analyst]

This is not an answer to the original question. Did you mean for it to be a comment somewhere else?

[Malini]

This is about the color segmentation , you mentioned

[Image Analyst]

Alright, I'll move it for you to where it should have been posted.