I try to model my house using pictures from each side. It is easy to scale one line to match the true length. However, images have significant perspective error.
Is there easy way modify the picture so that all house edges are scaled approximately right?
I believe that streching upper side of picture as much as vertical house edges becomes aligned, would be accurate enough. But couldn’t find a way to do it.
Any tips, what would be best approaches to use pictures with perspective errors?
CageEdit can be used to make parallel lines parallel. However it will also introduce distortion with the distances in different directions at different scales. You will need to scale the image in two directions at right angles to remove the distortion, using Scale1D twice or ScaleNU.
Or look for image processing software which includes “perspective distortion correction”.
Note: it is only possible to have distances at the same scale in a single plane. Anything in the photo out of that plane will be scaled differently.
I usually take pictures with my cell phone and found that taking pictures from a distance and sacrificing loss in resolution helps. If you can, use a digital SLR with the longest practical lens for the subject (reducing the distortion of the wider lens angles). You can’t shoot a tall building with a 120mm lens from close up…
Take your images as close to parallel and perpendicular as possible. If I were shooting a 2 story house, I’d get centered left to right and either hold the camera high or get on a ladder.
Like @davidcockey says, edit your photos to remove as much distortion as possible. crop sections into panels. For instance in photoshop your can perspective crop and modify layers with transform tools. Change the cropped image size to match the true proportions of the object.
now when you import the pictures you can scale a bit easier.
Thanks David! I dried before CageEdit and became frustrated for creating unnecessary distortion. I thought - there must be a better way to do it right.
Important note about “single plane”. I understand that. In my case house has only 4 planes - facades. If 1 facade has more than 1 plane at different distance, each must be processed individually.
Good point: maybe Rhino is not the best tool for manipulating pictures! (I should try Photoshop to see, if it does the job better!)
Very good point. Having longer telephoto lens helps a lot.
In some cases this can solve the issue and pictures are usable as they are!
Sometimes there are just not enough free space around the building and 1 or more facades cannot be photographed far enough.
I have also tried to use drone for obtaining pictures. I find it really helpful and easy to get the best possible pictures around the building at right level and perpendicular spot.
And if someone does not own a drone - maybe friendly neigbour will be happy to help!
In 2 minutes, I learned how to use “camera raw filter”,“convert to smart object” and correct the perspective automatically or manually with many controls.
Now I can handle any photo in matter of minutes - and continue in Rhinoceros with “good stuff”.
If you have Photoshop - it is really powerful, but not free. Maybe somebody can give a hint on how to do this with “open software”. I am sure that there are free solutions available.
Remeber that even with a long telephoto lens the scale of the object depends on the distance though the difference is less.
A shifted lens can be used so any planes parallel to the image sensor are not distorted with parallel lines parallel, etc. But there will still be a scale difference for objects in different planes. Unfortunately shift lens are not inexpensive and are not common. The exact same geometry can be achieved using software perpective correction as with a shift lens. The major advantage of the shift lens is the entire image can be used while software correction results in the resulting image not being rectangular, and will need to be cropped.
Photogrammetry is very different from using perspective corrected photos. It creates a 3D point cloud and/or mesh from a set of photos. The 3D point cloud/mesh can be imported into Rhino and used as a reference for modeling. I use photogrammetry with the images taken from the ground (using a tall tripod or sometimes a pole but not a drone) and processed in Metashape software to capture the shapes of boats.
Tilt/shift lens was new to me. Kind of interesting, how moving the lens off the center-line could help correcting perspective error. Looked at Ebay - and found some too. But for my purposes I am starting to appreciate good quality cell phone that is always in the pocket and compact drones.
Photogrammetry sounds really interesting and I hope to have time to study it more. One interest for me is to model churches from inside and outside. I think programming the flight path and shooting angles would help a great deal to gather enough data. Do you have a specific reason not to use drones? If your interest is small boats and the shape that is under water level - It is obvious.