I have a simple (10’ x 12’ shed roof) woodshed that I would like to photo match in order to build off it a car port.
I cannot imagine a good photo match of a rectangular structure involving less than four pictures. I took four pictures—clockwise—each from a successive corner of the shed. None of the photos were cropped.
Photo #1: I’ve done quite well in doing a photo match to photo #1 (no brainer). Positioned the axes origin (yellow square) as recommended in most photo match videos. Scaled to a spacing of two feet (all the measurements were great). Then drew and pulled a 3D model that accurately represented the shed.
Photo #2: Importing photo #2 went well, and realigning the axis required minimal work. I left the axes origin where I’d positioned it in photo #1. The model needed to be resized (click/move on an axis) in order to fit the photo, but it was a nice fit. [Going back to photo #1 confirmed that the model fit well.]
Photo #3: The import was straightforward; as was (very slightly) repositioning the (red, green) parallel axis lines. (To make the job easier I’d chalked marks around the shed). The axes origin (yellow square was positioned somewhere within the shed (indicating that it related to the bottom corner of the shed on the back corner (blind corner) of the shed. This time the shed model was the right shape (and the corner/sides pretty much lined up) but high (started above the bottom of the picture and extended that amount over the roof of the picture). All attempts to resize the model resulted in a model much too small to relate to the picture.
I doubt importing photo #4 is going to help, and can’t figure out what to do now. Something tells me that if there were the ability to (congruently) resize the photo I would be able to get the model to fit.
Photo #4: Nonetheless, I imported photo #4, repositioned the axes origin (to the location it was in photo #1); and aligned the red/green lines. The model was again the correct shape, but not the correct size.
Again, I think the problem could be solved by being able to resize photo#4.
What do you recommend I do? Also, are there any SketchUp videos or tutorials that cover importing more than two photos of the same structure, and dealing with model resize problems particular to those photos?
You don’t resize the photo, but, when setting the axes, you can also set the size of the measuring grid (by dragging along the axes) and also the size of your model in relation to the image and the scene.
I did set the size of the measuring grid (2 ft.) and for photos #1 and #2 (where I slightly resized the model) it came out fine.
Now it’s my understanding that the axes-origin SHOULD NOT be reset as additional photos are added. It should stay at the point where it was originally located. Is this right?
It’s also my understanding that the measuring grid should not be reset as additional photos are added. Is this also correct?
I think it is perfectly fine, and even necessary to move the axes when importing multiple “New Match Photo” scenes. The trick is to move it before the image is imported. I am talking about positioning the axes using the Axes tool, not moving the yellow-square axes intersection while in the middle of the MP setup. After the first image is imported and scaled correctly, I use the newly drawn geometries as a reference to size subsequent imports, rather than the grid.
I made a video about this!
If you want to jump to the part where I import a 2nd image and move the axes, it is at the 8:57 mark.
Great to hear this (the video is really helpful, and I haven’t gotten to the real complex (interiors) part.
Also, another aspect your video makes clear is how integral dealing with ‘photo texture’ while matching photos can be to the matching process (I never thought I would be texturing at all…just using the photos as ways to get the model accurate.)
Should I also assume that if realigning the yellow axes origin is OK (say because your original yellow square is no longer in the photo) that it is also kosher to rescale to the new photo?
(Part of the reason for this second reply is to see if I can get italic type to show up in a message.)
I am curious…
Why, given the fact the photograph is merely an artifact that is ‘traced’, and
given that the subject matter within a cropped photo is no different from that of the uncropped one (i.e., the lines still recede into vanishing points)
(Why) is it bad to crop photos when doing a photo merge?
I think it is OK, as long as it is a part of the setup. I like to think about the Match Photo process as reverse-engineering the spot from where the camera was when the shot was taken. Not only that, but the perspective lines allow for the focal length of the lens to be reverse engineered. While it is not necessarily procedural in that things need to happen in a certain order, I like to use the grid as the last step to reverse engineer the scale. After a few pieces are modeled, a ceiling height for example, I then use the model itself rather than the grid to match the scale of the photo to my virtual 3D environment. The photo is never really scaled, just the world and view relative to it.
Two potential issues to be aware of are this: If you scale (or move) the model after, and then return to the MP scene, it will no longer align to the imported MP photo. I used to skip the grid-aligning step and just use the measure tool to scale a known length. That is a bad practice, unless you absolutely no longer need the imported MP photo and it’s scene. The other thing to consider is that if the grid is moved while in MP edit mode, any textures projected from the image will remain from the view that they were projected from.
Think of it as comparing two camera locations, one in the real world and one in the virtual world. Moving perspective grid lines, axes locations (yellow box) and grid lines in the MP edit mode is really just moving the virtual world camera location to a spot that may or may not be closer to the relative position of the real world camera that took the photo.
Match Photo assumes that the center of the photograph is the direction that the camera is pointed. When a photo is cropped this is no longer the case. As a result the photo can’t be matched using just two vanishing points. We’ve discussed adding more features which could support cropped photos but this would complicate the UI. Maybe someday though…
