How to align image pixels
Aman Rana from United States  [25 posts]
1 year
Hi there
I have 5 images capture using multispcetral camera at UAV. ie blue, green, red, NIR, and Red edge respectively. I am really finding hard to align all image so that I can work on pixels level. Now each pixel is not have same ground information due to vision disparity as each image is captured by different camera. I want to use math on pixels value of these images to identify vegetation like NDVI or WDI etc. I will appreciate if anybody come up with any suggestion or solution.

Steven Gentner from United States  [260 posts] 1 year

This is possible to do with a couple tricks. The first is to note that because each camera is essentially offset in a know direction from the other the alignment search can be reduced to searching along the disparity line between any two cameras. While the disparity amount will change with distance (in this case height of the UAV) the correlation of the two images will always be along the same angle. This just helps to reduce the computation when comparing the two images.

The biggest issue is that because each image is a different 'color' you cannot just use the ordinary stereo or alignment modules in RR as they assume the two images contain information that is to be matched with the same intensity. In your case, what you are looking for is quite a different match calculation where you are looking to match intensity A pixels in one image with as many intensity B pixels in the other. Since they are images of the same content the formula B-A should be most stable when the two images are aligned. So it is possible to do (not sure if I'm explaining it well enough) but it is a custom solution and possibly much more complex than described since I didn't take into account any tilting of the AUV during a capture. Currently we don't have a module specific to this type of correlation.

Naturally, if you can ensure that the AUV always flys at the same height and is level to the ground at the time of a capture, simply specifying the X and Y offsets from one image to the next will get you what you need too.


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