Help with edge detection - surface tracking program

Hey Everyone,

Thanks in advance for your help. I am working with videos where I track a moving surface over time. Depending on what the surface is made of, it will move slightly differently. Different materials will allow different amounts of light through, or will make streaks that make it difficult for me (a novice) to track the surface. I have written a surface tracking code that is pretty mess, and works well for "good" images. Here is a link to a good image and the green lines shows the surface that my code tracked http://www.flickr.com/photos/43987619@N04/6994389448/. The following image is a frame of a video that I am having trouble tracking, here is the image http://www.flickr.com/photos/43987619@N04/6994372858/ , and here in this image I have drawn a red line to show the surface I would like to track, http://www.flickr.com/photos/43987619@N04/7140478777/ . The other type of problem image I have is given in the following link, http://www.flickr.com/photos/43987619@N04/6994389486/ and the same image is shown with a red line to describe the surface I am trying to track http://www.flickr.com/photos/43987619@N04/7140478791/ .

I won't post the code I am working with, because I think it is messy, hard to explain, and probably not very good. For this I have been using the canny edge detection, wiener2 filter, and adjusting the intensity of the image to try and help get rid of noise.

I hope someone can help me make my code more robust since I spend most of my time changing the intensities for each video, and then again for certain frames inside of the video. If there is a better way to share images, I can do that.

[Manisha Singh]

Hi,

I have similar problem. Did you find any solution

Answer 1

[Image Analyst]

I think you have an image capture problem more than an analysis problem. Your analysis would be so much easier if you improved your image capture situation. It looks like you have a light source staring directly down your optic axis and that is causing a lot of glare and internal lens reflections. Try moving your light source off axis, or behind the camera or above, or use cross polarizers or something. Those images are fairly poor and I find it hard to believe they can't be improved, which would make your analysis a lot easier. I never would have guessed at the red line on the last image so I don't think I'd be able to design an algorithm to find it if even I can't see it.

[Casey]

Thanks for the help. I have been working a lot with the image capture. I put a polarizing filter on the camera, but it doesn't seem to help much. The camera is a HDR camera, not sure if that means much to you or if it is helping (or hindering) my image capture. When I purchased the camera, it was the best cost for frame rate and resolution. I am limited in light placement. Here is an image of the experimental set up. http://www.flickr.com/photos/43987619@N04/7141777067/ As you can see, we are tight on space. If I put the angle the lights directly downward, you can't make out the bottom of the surface as it gets deeper. I wonder if I turn off the HDR if my images will get better.

Casey