Seam tracking
Nilanjan from India  [4 posts]
13 year
Dear Steven
I am enclosing a jpg. I need to find the co-ordinates of the point of intersection between the laser line and the vertical seam. I tried sobel edge- vertical and horizontal, not able to achieve results. Can u pls help?

Anonymous 13 year
Two things you can try ... 1. attached is a simple robofile that uses the fast harris point detection. The laser light against the seam makes for a very sharp black/white transition which can be detected. My concern is how stable this feature is on other images.

2. If you have access to the camera tune up the shutter speed which will make the image much more dimmer but will keep the laser light much brigher than the rest of the image. Laser lights always saturate the CCD and thus increasing the shutter speed will remove other white light from the scene but keep the laser light ... if you can do that it is much easier to process just the laser light to find the dip in the line. You will also need to use the radial undistort module to correct for the rounded warp of the image.

If you have more images and/or adjusted images with faster shutter speed please post them here and we can continue to check on techniques that maybe more reliable than the attached.
Nilanjan from India  [4 posts] 13 year
Dear Steven
I don't have a method for adjusting shutter speed. You were right, I have situations where Harris corners are not stable. Enclosing a sample jpg

Anonymous 13 year

Ok, but it appears that you do have control over the position of the laser and the camera? Is that correct?

The desired effect of increasing the shutter speed can be seen in your currrent image where the laser light is placed in the darker area of the image. This brings out the laser light without the reflection being in the middle of the laser line as seen in your first image.  

Attached is another technique. Basically it eliminates everything from the image except for the laser line (straightest largest object in the image) and then orients the line to be as straight as possible. With this is just detects the lowest point of the line which will be the seam dip in the line.

Note that we use the radial distortion module to remove the warping in the image ... you may want to play with that module a little more to get the values that produce the straightest laser line.

Again, more images to test the better. There are many ways to do this but only testing on images will reveal the best. It is also recommended to use a tripod so that you can take images in a reliable and standardized way. For example, if you can ensure that the laser line and the image are straight then the orient module would no longer be needed.

Note that this assumes that the laser line is somewhat isolated from any other white light source ...

FYI, your image is include in the robofile so it should just run as is.


This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index