Photo-match problem

DIDEC

Hy!
My first post I have drawn a design for a new playground for a school. Buth I have a big problem using the photomatch in sketchup! Normally this works fine, buth this time I really dont get the perspective right. I did notice the 2 photographs have also a weird perspective, I think they were taken with a wide angle lens. Is it possible that sketchup cant match this? Pleace take a look at this for me if you think you can work with the photomatch. I added the 3d file, the 2 photos and the ground plan in the zip file...
Thanks for helping!
bye
http://www.megaupload.com/?d=54U5DG67

mitcorb

One suggestion/question, not sure if it is correct, did you set the camera FOV field of view?

DIDEC

Thanks for the quick reply!
I did play a bit with the FOV, buth if you work in photomatch, doesnt sketchup adjusts the fov itself? I did some searching on the forum on found another workaround which I am trying right now: inserting the image as watermark and manually trying to match the perspective...
Bye

The Pixel Artist

(I originally posted this on the Thearender forums, but thought I'd post this here too. Maybe it'll spur some good discussion)

Hi Didec,

Since it looks like nobody has been able to help you yet (saw that you posted on Sketchucation too), I thought I'd try to help.

Yeah, for whatever reason these are tricky images (they shouldn't be!), probably due to the wide angle. I have this problem all the time too, so it's not just you .

Unfortunately, PM (Photo match) is just fundamentally flawed in design (and V8 has actually made it worse!). Using the PM tools as they were designed keeps getting the focal length wrong (finding something like 48mm). I personally wish it would just read the photos EXIF data and/or give us the options in the PM dialog to manually add focal length, aperture, and adjustments for things like barrel distortion. If it did, it would it find the equivalent FL should be 27.4mm to match the 17mm you used on your EOS camera.

Anyhow (rant over) I was able to get an approximate match for the two images by fudging the heck out of the perspective controls and horizon adjustment nodes. They're pretty close... but not perfect. You can probably spend some more time tweaking what I've got and get it a bit closer. Sadly, if I need something more accurate, I usually have to use another app like 3ds Max (but then you need good height data too).

Best of luck,
Adrian Alan Brown

LINK: http://www.megaupload.com/?d=9IYHRH3V

DIDEC

Thank you so much for checking this. I am glad it's a problem with sketchup and it's not just me sucking in this photomatching I gave it a try when using the pics as watermark and this worked quite well actually! I saw your photomatch and my matching were really close, so thats good news. I posted my final setups, so if you want to check them out you can compare. Now exporting to thea and lets see what it gives. I'll post my results in a few days!
BTW one question for the future, what would be the best to do when taking pictures for photomatching? Do I have to use a standard 50mm lens? Some other things to keep in mind to make good and easy use of sketchup's photo-matching?
Thanks again for your help!
Bye
http://www.megaupload.com/?d=XO356CVD

The Pixel Artist

Hey Didec, no problem. Just hoping to help.

Yeah, that watermark technique works pretty well and generally use it more than I do PM (unless I need the photo for texturing). I actually had to used it about a week ago when a client sent me a camera photo that I needed to integrate a rendering into. (Here's that image: http://www.thepixelartist.com/c2006008/p2010011/20100709/p2010011-20100709-view1-1600.jpg )

But, it's a bit of an inaccurate and tedious process to get right. Which is a shame, because that's the whole point of having PM. Plus you lose the option to use the photo as a texture. What's strange is that PM seemed to work better back in V6 when it first came out and has seemed more glitchy since V7. For example, here's an animation I did back in V6 where I used PM to model a very complicated existing building. Used a lot of photos, wide angles and mixed orientations. I haven't had nearly as much success with PM since then (LINK: http://www.thepixelartist.com/c2006001/p2007011/20070525/test1.html )

V8 has made things worse by adding the front image option, but then removing the ability to use display controls, forcing you to go into the style manager to change the opacity settings. It's really pain in the ass now! All they had to do was add those controls to the PM dialog box (opacity sliders for both front and back image, plus keeping the display controls). Frankly, PM was just very poorly implemented in the first place and just needs a major overhaul with a far more accurate/precise tools... and could be made far more intuitive at the same time (like most of SU really... but don't get me started! )

Anyhow, to offer a bit of advice on how I usually handle a photo integration project (if it needs to be fairly accurate), I usually start off with an aerial print of the site (obviously off google or bing), pre planning my photo location as best possible, then accurately mark the actual photo location out on the site (along with tripod height per photo). If needed, I'll even measure to the center of the tripod location with a 100' tape to the two nearest objects most visible in the aerial. Also note your focal length/aperture (or just get it from the image EXIF data). If possible, I try to set my camera perfectly level so I don't have to worry about pitch and roll later (I actually use a bubble level attachment on my cameras hotshoe mount).

From there I'll bring the aerial data into SU (or ACAD). This is one area that V8 will certainly be helpful since we can now bring in much higher res, color aerial data. Before I would have to export the lower res, B&W aerial texture from SU into photoshop and update the texture with highres color data manually.

Using the "Film & Stage" plugin (which I think is just built in now) I will make a camera object and manually place it on the aerial at the correct location and height. I'll use the "Edit Camera" option later to change the cameras focal length and aspect ratio (typically 1.5, like your camera, or 1.3333).

Now, this part is a little hard to explain. I will build a triangle (around 10') from the center of the camera location, flat to the aerial, with its main angle to match the "horizontal field of view" of the photo. For example, a 28mm FL would have a horz. field of view of 65.5deg (Didec, your 17mm EOS 550D would be 67). Here's a great online calculator I use, use "Angular Field of View" link: http://www.tawbaware.com/maxlyons/calc.htm. I add construction lines to the main angle edges and rotate the camera to bisect the angle. You can make a group of the whole thing, then rotate it (around the location center) to match the view in the match photo (align to the objects at the edges of the photo). There's a couple more accurate ways to align the angle, but you can usually just eyeball it. I've attached an image to hopefully show this step a bit better.

Add the photo as a background image (again, make sure your aspect ratio is correct) and it should be pretty close. You can tweak the camera location manually if needed. I usaully use the "Walk" tool to keep the camera location set.

Here's a couple images I've done using this technique.
Before: http://www.thepixelartist.com/c2007002/p2007013/20070625/p2007013_20070622_view1-1000-exst.jpg
After: http://www.thepixelartist.com/c2007002/p2007013/20070625/p2007013_20070622_view1-1000wm.jpg
Before: http://www.thepixelartist.com/c2007002/p2007013/20070625/p2007013_20070622_view2-1000-exst.jpg
After: http://www.thepixelartist.com/c2007002/p2007013/20070625/p2007013_20070622_view2-1000wm.jpg

Hope this helps. Best of luck!
Adrian Alan Brown