[Asking for solution! No code!] Column detection with/without stereo vision
Hello guys, I am sorry to ask without code. I tried to calibrate my stereo fisheyes and failed, tried for a week to do it. Rectification and stereo calibrate is bad somehow. Anyways, my plan is for the drones. Drone will have one or two cameras and it will fly on a vineyard. As you know, vineyard is basically in column shape. So when flying over a column, I need to understand that I am in the middle of the column . Otherwise it means that I need to go left or right. Example:
If you can give me some hints I'd be so grateful. I think that if I had depth image I can take gradients to see if my rotation and location is correct or not. Or I can do some segmentation and split the images and get different gradients. It is so confusing. I don't want code, I just want some ideas. Thanks to all, great day.
Edit 1:
I am adding colored undistorted pairs and grayscale undistorted pairs with disparity(because of failed stereo calibration and rectification):
please put your image here , not on dropbox.
Can't do, everytime I upload image it doesn't add it. Sorry. I made it public so at least people can see the images.
resaved as jpg, and uploaded.
(there's a 1mb limit for images on this site, so please also respect other ppls bandwidth)
I didn't know that. Thanks for editing, I'm grateful.
to get back on topic, - what's the reason to use fisheye lenses here ?
(i doubt, that you need a real large fov for this task, and the distortions clearly get in the way.)
also, even if you can't do a proper stereo calibration (yea, hard !) , you should still try to calibrate the single cameras, so you can undistort the images
My drone should see the down and the front at the same time for landing(with aruco) and also I need to use it for column detection. I can't stereo calibrate but I did it single and undistorted the image. But even if I do that, it is a mystery for me to solve the rest. Fisheye for wide angle for both bottom and front that's all. Other than there is no reason. I am editing the post so you can see the undistorted images.