Line-tracking/3-point tracking

object-tracking

#1

Dear all,
I need to track two solid objects that are moving relative to eachother. I need to line-track (I have three lines to track). How can I do that under fiji/imagej?


#2

An example image would be helpful to get better answers.

Generally you would want to:

  1. Segment the lines by running a line detector ( e.g. https://imagej.net/Ridge_Detection )
  2. Select some point of those lines that you want to follow (e.g. the center of the line, maybe the endpoints)
  3. Get that point into TrackMate ( https://imagej.net/TrackMate )

Honestly, this is not going to be straightforward, but it is possible.


#3

Thank you for the reply. I have attached an example.


#4

Thanks for the image
img

Now the question arises what moves and how does it move.

Please provide at least two consecutive images and explain in detail what you like to obtain.

Regards

Herbie


#5

Thanks for your attention. I have attached two consecutive pictures, now and also the lines that I want to track(I have shown you those lines to be tracked in the overview).


#6

Your images arrived savely,
but I must admit that I can’t detect any shifts in x and y of the indicated borders in both images.

My approach assumes that there are no rotations and no scale changes and that the whole object is rigid.

Could you please check if there is actually an xy-shift between the two images of at least one pixel?

Regarding beam hardening, it depends on the X-ray spectrum and the material you are investigating. I’m no specialist in this field and you should consult thhe person who is responsible for all scientific X-ray investigations that are performed in your laboartory.

Regards

Herbie


#7

Thank you so much for your reply. There is certainly no scale changes as far as I can see and yes, the objects (I have assumed there are two seperate objects here as a safe starting point) are rigid.
I would like to do denoising and line/3-point tracking iteratively. could you give me an idea of the plugins for these tasks that might best suit these tasks?
thanks for your attention


#8

Sorry but my central question has not been answered (I could not detect any shifts):

Could you please check if there is actually an xy -shift between the two images of at least one pixel?

Can there be rotations?
Can both objects move independetly?

I would like to do denoising and line/3-point tracking iteratively. could you give me an idea of the plugins for these tasks that might best suit these tasks?

Why would you want to do so?

Please concentrate on my questions, otherwise we need more time than necessary and this is not what I’m willing to do!

Regards

Herbie


#9

Herbie,
Thanks a lot for your reply. I am sorry I didn’t pay attention to your question last time.
I am not sure if I understand how to mesure if there is an xy-shift between the two images (or amongst all the tifs as I have a set of over a thousand tifs of the kind).

regarding possible rotations or whether they are independent; that is what I want to find out by carefully tracking these lines.

I thank you again for your time.


#10

Sorry but your problem appears not very well-defined.

You didn’t answer my question regarding:

I would like to do denoising and line/3-point tracking iteratively. could you give me an idea of the plugins for these tasks that might best suit these tasks?

Why would you want to do so?


The approach you need for tracking depends on the kind of movement you’re expecting. If there are neither rotations nor scale-changes and if there is only one moving object, then cross-correlation willl do without any pre-processing or denoising. (I tried this with your images.)

However, cross-correlation doesn’t work if two objects move differently.
If this is the case you need to look for more involved approaches that most likely will require pre-processing and denoising.

Good luck

Herbie


#11

Thank you Herbie,
I will speak with the experiements responsible in the lab, to get a clearer perspective of the behaviour of the object(s). However, since I am a still a beginner in ImageJ, I would be grateful if you could explain what kind of approaches you would use for pre-processing and denoising. (for example, I did not know what cross-correlation was, i just looked it up).

I would thank you anyway for your attention.
Ati


#12

Denoising always has an impact on the spatial resolution (sharpness) of an image. The simplest way to reduce noise is by low-pass filtering.

If you are a beginner I’ld assume that you’ve already studied the ImageJ User Guide.

Regards

Herbie


#13

Thank you. Yes, I have been looking for eductional videos and I am going through the documents. /Ati


#14

From your picture it seems you have two kinds of lines. The more horizontal lines look like edges where it is dark on one side and light on the other. For that it would use an edge detector like https://imagej.nih.gov/ij/plugins/canny/index.html . You do not need to interatively denoise and detect because both are integrated.

The vertical line looks like a dark ridge and for that you may want to use the ridge detector I mentioned above. You may need to invert your image first to make it a bright ridge. The edge detector may also detect the sides of the dark line.


#15

Thank you for your comment as well. I am trying this approach. I have added it to my plugins. But as I am a novice in using ImageJ, I am still gong through the documets and I will have to speak with the lab person as well.


#16

Hello Atha,

There is no movement between the images of the objects you are imaging, but there is substantial differences in the noise of the two images. They also have an unusual spectrum. Would you possibly send us some more info on how the images were acquired?

Very curious,

Bob


#17

Thanks for confirming Bob:

I must admit that I can’t detect any shifts in x and y of the indicated borders in both images.

Herbie


#18

Hello Atha,
High Energy imaging (xuv,x-ray, gamma and beyond) is different than Lower energy imaging in many ways. I am sure you have probably noticed in being a Materials Scientist, so I could not help but be very curious as to the cause of the pattern of the noise difference between the two images you submitted. Not only in intensity and quantization of intensity but in the fact that it seems to eminate from the areas you wanted to detect movement in.
Although there was no discernable movement in the actual objects, there was a substancial change in noise pattern which you might find interesting to investigate. I don’t know what it was you are actually looking for nor the exact way of aquisition so I can only theorize the why. If the information is propriatary I understand but I would like to know what you could tell me of the aquisition process.
I hope you can cure my curiosity.
Bob


#19

Hello back,

Thank toy for your comment. I cannot disclose anything about the images or the acquisition at the moment. I have recently started to work with Imagej and was curious about what I can do to get a meaningful result (movement detection) on such stacks of images.


#20

Hello again,

I would appreciate it if you could suggest a methodology of denoising for the kind of images I got.
Thanks in adance for your attention, Atha