I just started using and learning Sketchup Matchphoto and everything was going well until I added a second picture. When I added a second picture I noticed ALL of my lines from the first were all messed up in terms of where they are supposed to be by up to a foot on a 40ft long building which means tracing the rest of my lines is impossible.
I watched some tutorial videos on it from SketchUp and the person in the video just flawlessly drew their lines all of the place and came out with a perfect model, yet try it and everything is totally wacked.
I need to use it to transfer somewhat accurate buildings to 3DSMax but I can’t as long as this happens. The images haven’t been tampered with either and I set the points relatively where they should be.
Has the one you’re having trouble been cropped or otherwise manipulated? For Match Photo to work, the photo can’t be cropped or have any perspective correction done to it.
Neither of those two images have enough useful lines for setting up Match Photo properly. The best you could do is eyeball it but I think you’d be better off to take a few measurements and draw the shed without using Match Photo.
I had not played with multiple images for match photo before. My first thought was that perhaps both images have to be the same scale, so I trimmed both images to the shed, and then scaled both bitmaps to be the same height. That didn’t help. Setting up the second image and making a model ends up not matching the first image at all.
Not sure what the best way to work is, other than model just based on one view, then use the other view to texture the opposite faces.
As @DaveRSaid… The bottom of the shed is not visible on either photo (ended up guessing where to put the green line based on when the blue lines were vertical). Also, why are you trying to line up Photo-match twice? On a simple shape like this, you can get the entire shape using just a single photo (or, as Dave pointed out, a half dozen dimensions).
It is entirely possible to use multiple images for Match Photo with a single model. You have to change the model axes before adding the second image. It is useful if you have multiple good images and you need to use the images for textures. But the operative phrase is good images. That means they have defined lines running off to the left and right vanishing points and it’s best if there are some high and some low in each photo.
If I start with one photo, do red and green lines, Done, make some models, add another photo, set red and green lines, the model is nowhere near the same. I used the tape measure, and the height of the two views, that had the same size images, are quite different. The first view model on the second view looks about half size.
This is exactly the problem I am confused about, the images ended of being lined up correctly just like the other side yet after I draw the lines on one side and move to the other it just isn’t even similar making the model not even square anymore.
@DaveR I came to the same conclusion of just making my lines for two sides, then finishing off the foundation and mirroring that for the roof making a perfectly dimensioned building like it is IRL. Problem is I still have those doors and gutters indented and sticking out of the building that need to be traced and modeled and it is hard to do that without having the other pictures being setup correctly.
One thing I need to figure out is how I can copy and past, or offset, say, the short wall in the picture to the opposite side of the building so it is perfectly proportioned. Do you know how to do this?
Then I should be able to easily add the roof and work on doors.
The main thing I am concerned about is if this is going to happen continuously, because right now I just went outside my home and took some pictures to practice on, I am afraid that when I go to get pictures for a real project that I will run into this same problem of not having pictures lining up with my lines correctly.
Faces have two sides, and SketchUp by itself has no problems with either, but, for instance, when you use your model for rendering, many rendering applications either ignore the back sides of faces completely or refuse to apply materials to them, or show them as black.
BTW, I think that a 2.5% error using such a simple photogrammetrical process is rather small. In addition to the factors pointed out that make the setting up PhotoMatch difficult with your images, there are a bunch of accumulating sources of error, for instance, all lenses, especially zoom lenses, have distortion, and PhotoMatch has no calibration for individual lenses, and all cameras, especially if you shoot JPG, already make corrections to the image in-camera, so for the most part a non-manipulated image doesn’t exist, and it would look distorted.
I have seen many sets of measured drawings made by specialized firms with laser measuring equipment, and for the most part even those have obvious errors (a recurring one is that the measuring tolerances make prefabricated concrete columns in a building display odd lozenge- or trapezoid- shaped sections)
Hey Anssi thanks for the reply. I think one factor is that I may have changed the zoom on my camera, I need to go take some more pictures and then see if it is corrected if they are all at the same zoom level.
How does anyone get accurate dimensions with photomatch then? They don’t have to be exactly right like from a blueprint for what I am doing but if I can get the images and lines aligned exactly or extremely close then it will save me probably dozens of hours as I plan on using it for many, many buildings.
Zoom level isn’t very critical but wide angle shots can be difficult because of fish eye distortions that can occur.
They start with decent photographs that have suitable lines running off to vanishing points on the horizon and which have not be cropped or otherwise manipulated.
