Motion Detection in Multiple Locations

[dragonlordged]

My project involves keeping track of motion in several discrete (but close by) locations on a glass plate etched with a specific pattern. Unfortunately, the locations which need to be watched for motion are in relation to the pattern on the plate instead of to the camera. My first idea was to use image analysis tools to find the edges of the glass plate and re-orient the image based on that, but I have been unable to find or create an algorithm with which to do that. Additionally, I am concerned as to whether the transformation of the image would be possible within my time frame (as I would have to transform each frame and check for motion in it and wish to have a framerate of around 10 fps, though I should be able to handle less). If anyone has any suggestions, I would be happy to hear them.

[Vivek_N]

Is there any reason why image analysis is not working for you? If you are coming across any specific problems with your vision code, you can post your questions here.

If not using a vision algorithm to keep track of these points, it would have to be through some type of sensor attached to these points such as an accelerometer. While we provide DAQ cards that are capable of reading the values obtained from the accelerometers, the algorithm to correlate this to position would still have to be architected. 

[muks]

Can you also post a few images?

[dragonlordged]

Image analysis isn't working because I can't set up static locations to monitor for motion without making sure the image is always aligned the same way (which it won't be, due to small movement of the physical object) and I am unable to find the correct edges in the (rather dark) image in order to realign it so I can set static locations for monitoring. I have tried thresholding the image and then performing edge detection, but it is still not working correctly.

A sample image is attached.

dlg

[dragonlordged]

Actually, the labview vision algorithms seem to work only with b&w images, so we have been taking the images in greyscale, not in color. I do not have easy access to a black and white version of the above image, so you will have to use your imaginiation.

[muks]

---

the labview vision algorithms seem to work only with b&w images,

---

I dont understand what you mean. You can apply a lot of greyscale filtering functions. Also in the image you posted did you try applying gamma correction and edge detection?

[muks]

Run the vi attached with the image u attached.

Message Edited by muks on 01-21-2010 03:33 PM

[dragonlordged]

Thanks muks, that's a very useful VI. After fiddling with the contrast and color extraction, I managed to make it do basically what I want it to do. I'm just wondering, where did you find that vertical edge VI? I can't find any documentation on it and am also looking for a find horizontal edge VI for similar uses, as well as wanting to understand the edge detection VI better.

dlg

[muks]

dlg,

I suggest you first play around with vision assistant. then you generate the vi from it and you can always optimize it later in labVIEW.

[muks]

in lv