Match Photo Field Of View
-
I am new to the forums and have been reading about problems with match photo and field of view in the search engines.
I have modelled a loco shed complex from 1 2500 scale maps and am trying to match photo from a 1966 colour slide, but match photo always loads and alters to a set camera angle and also moves my viewpoint.
Therefore, variable match photo field of view is something that I would wish to see in sketchup 9.0
kind regards
Nick
Bristol
Uk -
I apologize if I don't fully understand what you are getting at. Photo match is only meant for you to build a model based on a photo. If you built the model from plans then you need to use a different procedure as photo match is not the tool to use.
Step 1. You need to figure out the camera viewpoint in your photo. Where exactly was the photographer standing and how high above ground level was the camera.
Step 2. You need to set the camera position in your model to match the position of the original photographer. Then you can cut out your model and drop it onto the photo.The only potential problem with this is you are stuck with the original photographer's camera position. If his camera position sucks, then your render will suck.
Actually it is possible to get around this problem but the solution may not be practical. It would involve two models. You model the whole world seen in the original photo and then integrate it into the model of the loco shed that you have just finished. However depending on the background, that could be a lot of complex modeling.
Would you mind posting both the photo and your model. I would like to study both in a little more detail.
-
Nick. I don't know if your photo shows the way the site is today or if it is an old historical photo. If the photo is modern, your can set up your rendered view for best composition and angle and then figure out where you would have to stand to get a matching photo. Once done, you can go back to the RR yard and take a new photo to match your model. Of course if it is a historical photo and the yard has been torn down, you are out of luck.
-
Well, PM can also be used to reconstruct a certain (here, say "historic") camera view for an existing model one made from plans. However in such a case, the idea itself is that you match the camera view of your current model with that of the photo for some purpose (say to render a SU output that you can compare with what' on the photo...)
Think of architects designing a new building in an urban environment; they have the new design and then match it with a couple of shots taken on site so that clients can see how a new building will look like in its environment.
However I also do not fully understand the point here. Your match photo scene will almost certainly have a different camera view than what your current is. Just add a new scene to preserve it. Since PM has the sole purpose to get the camera view of a certain photo, it is natural that it will "change" your view. You cannot alter the view of an existing photo.
So I feel we are not in the same boat here - please, explain where I misunderstand you.
-
Hi Gaieus
I would love to know how you can use photo match to match a camera viewpoint from within an existing SU model? -
Simon and others,
I don't have any answers for you but have struggeled with the same dilemna. It's been mostly trial and error, even after reading some articles on digital camera focal lengths and their relationship to a standard focal length.
My digital camera's EXIF data, indicates a 6mm focal length. I've read that to convert it to "standard" focal length numbers, you basically multiply the digital camera focal length by 6. So, my 6mm is converted into 36mm. I thought then, I could use that number and plug it in to SU, which should superimpose nicely on my photo, but it never works out. It seems more often than not, that I need to bump up my SU focal length to almost 60mm to get the same "feel" as my photo.
Anyone know any other tricks, if there is a better conversion or if I'm missing a step?
Ben
-
Ben, photo matching has absolutely nothing to do with focal length. The position of your virtual camera in SU has to match the same point where the real camera is or was positioned at time of exposure. That's it, doesn't matter if it is a 6mm lens or a 6000mm lens.
-
Roger, I understand that one of the goals is to get the camera position to be the same in both the model and the photo I have, but are you saying that changing the focal length setting in the SU doesn't make a difference if I'm trying to overlay/insert a model into the background of an existing photo? It seems to make a difference in how much the perspective lines converg.
Thanks.
-
No, IMO Roger wanted to say that what does not matter is that what kind of camera/lens you took the photo. Theoretically two, different lenses could produce the same photo in the interval they both cover.
-
Gaieus and Roger,
Didn't mean to hijack Nick's post, but I'm not sure I follow, or perhaps I didn't explain myself well enough.
My question is, I've taken a photo and want to insert, you name it, a new building, etc., that I've modeled in SU, into that photo and make the perspective look right. Other than, getting the camera location to be the same in my model as it is in the photo, what else matters, the cameral lens I take? The focal length setting on the camera? I understand Gaieus' comment to mean it does matter what type lens I shoot with, or perhaps a setting on the digital camera. Perhaps I just don't understand focal length and lens types, that's making me no understand.
I appreciate each of your willingness to help. Sorry if I'm slow to comprehend.
-
Have you seen this http://www.youtube.com/watch?v=MTFQrJQlByA ?
I watched it but I don't recall the exact method. They do take an existing model and put it into a photo. I believe you need to model something in the photo view that gives you an entity reference in space and scale for positioning the building. Say there is a fence from which the building is to be 5' distance. Then you can draw lines to the building corner.
Imagine if you MatchPhoto/modeled a building from the photo and put a couch from your collection on the porch. You are doing something similar except the couch is a whole building. You need something modeled from the photo to start.
The nature of MatchPhoto doesn't promise perfection or an ironclad route. I've had some success using photos for the model background via Match Photo (as opposed to modelling whole existing structures), but I usually did most of the modeling after the photo. Hope this helps.
-
Ben, perspective is determined by lens to subject distance and not camera focal length. If film or digital sensors had infinite resolution, all cameras would only use one lens. You would just take a picture with the widest angle lens you could buy and crop into the picture for the composition you want. To demonstrate this. photograph a building with a telephoto lens. Then photograph it again with a wide angle lens from the IDENTICAL spot. Of course, the building will be tiny in the wide angle photo. Now enlarge the wide angle image of the building until it is the same size as the telephoto shot. Using transparent layers in photoshop, put the enlarged wide angle photo over the telephoto picture. The two images will align perfectly with the only problem being that the enlarged shot will be grainy (if you used film) or pixelated (if you used digital). The point is that if both the wide angle and telephoto shots were taken from precisely the same point, the perspective (convergence) will match precisely.
Now to drop a building into an existing photo, your view of the building will be controlled by where the original photo was taken. If it is a bad view point for rendering the building it is tough luck if you are using an existing/historical photo. However, if you can use a new photo, you can angle the render to your desire, figure out the precise SU camera viewpoint and take your real world camera to the building site and set it up to match the SU view point. If you do it by this second method, then lenses will make a difference because your view point (lens to subject) will be determined by your render and you will have to change or zoom your lens to match the fieled of view of your render.
Something to keep in mind when doing the render first is the height of your viewpoint. If you set it too high, you will be standing on ladders, dangling from cranes or hanging out the doors of helicopters to match the renders viewpoint. I have done all these and it can get scarey so try to set the eye height in the render close to your own eye height.
For more detailed info on setting camera position in SU see:
http://sketchup.google.com/support/bin/answer.py?answer=95028Most people have a resistance to believing what I just explained. I know I did not believe it even though I had done perspective renderings in both architecture and industrial design classes. However after the subject came up in my Army Photo Officer's class, the instructor sent us into the field and we did the whole exercise. Much to my amazement, if the subject to distance matched, so did the perspective.
-
Peter and Roger,
Thanks for your help.
What I've done, and why I'm asking, is taken a photo first (so I know I reach the location again), then modeled what I want to insert. But I also modeled the context of the existing information (buildings included)that I shot with the camera. Now I go back into SU, place my camera location roughly where I took the photo from and insert this modeled object into the photo. So the distances Roger spoke of, theoretically should match in both. But somehow the perspective lines aren't that close. I don't usually have a problem with getting the scale of the objects to match, just the perspective lines. I've done this on more than one project and for all practical purposes, get close enough that it's hard to tell any error, I've just been trying to figure out how to get even more precise than trial and error.
Again, thanks for your help. I'll watch the videos at some point too.
Ben
-
OK Ben, I guess I skipped some crucial detail. The amount of virtual camera tilt must also match the up and down tilt of the real camera. Could you post your photomatch project so I can see where things seem to be going wrong?
Advertisement