image-based 3D modelling

[Devin Johnston]

Is anyone using this software?

http://www.photomodeler.com/applications/architecture_and_preservation.htm

 

It looks very useful but I'm unsure if it's the best software out there, I'd appreciate any input on this and if you have other software solutions in this area please post them.

[Dave Buchhofer]

have you checked out realviz's imagemodeler? If you have access to an autodesk subscription plan they recently released it on there to subscribers for free.

[ctk111]

My office has a copy. What sort of project or you looking to use it for?

[Ernest Burden III]

I've always wanted to get Photomodeler, but never needed it badly enough to justify the cost.

 

I do sometimes do modeling from camera-projection, where I use a Cinema4D plug-in to determine the location of a camera and then start putting in planes and such, building up a scene. It's not all that accurate in the end, but for backgrounds it works fine.

[Devin Johnston]

have you checked out realviz's imagemodeler? If you have access to an autodesk subscription plan they recently released it on there to subscribers for free.

 

I'm in the process of checking it out, do you use it?

[d_mccreadie]

Hi Devin,

 

What is it you are trying to do?

 

I work with a company that are the agents for photomodeler in the UK. They mostly use Photomodeler as a low cost way to measure from photographs but are now using Photomodeler Scanner to create point clouds for accurate moddeling and measuring.

 

1st is a carved face that was photographed using a camera phone and then Scanner was used to create the 3d model and it was printed out on the 3d printer. This was done as an example to show that almost any camera can be used.

 

2nd image is the Palladio statue in Chiswick Gardens. The statue is being monitored for erosion using photomodeler scanner.

[Devin Johnston]

All I want to do right now is create context models of buildings or interior spaces, I know Image Modeler & Photo Modeler can do this but since I already have a copy of IM I suppose I'll give that a try before looking at PM.

[Chad Warner]

Maxer-

 

Let us know how it turns out. I'd be interested to see your results.

[Devin Johnston]

So far I haven't had much luck with ImageModeler, I'm having the same problem that I had with Match Mover which is the coordinate system. The process seems strait forward by creating location points on a picture you are defining the 3D surfaces of the object, when you get enough of them the image is calibrated and a coordinate system can be created. Problem is that every time I create one it doesn't line up with the buildings right angles. I've gone over and over the tutorials (which suck by the way) but just can't get it.

[quizzy]

there is a few do's and dont's in IM, but everything else is very straightforward.

during the calibration process you add markers, as soon as you are going to model in 3d, dont add any markers.

use imagemodeler 2009 and be sure to download the sp1 for it.

 

edit: Ow yeah.. here is an example of a city I modelled with IM:

http://www.3idee.nl/google

[Devin Johnston]

edit: Ow yeah.. here is an example of a city I modelled with IM:

http://www.3idee.nl/google

How many buildings did you model and how long did it take?

[William Alexander]

So far I haven't had much luck with ImageModeler, I'm having the same problem that I had with Match Mover which is the coordinate system. The process seems strait forward by creating location points on a picture you are defining the 3D surfaces of the object, when you get enough of them the image is calibrated and a coordinate system can be created. Problem is that every time I create one it doesn't line up with the buildings right angles. I've gone over and over the tutorials (which suck by the way) but just can't get it.

 

IM requires a little finesse and/or not trying real hard for perfection. You can but it will drive you crazy.

 

I don't know how many images you are using and how those images where taken. If at all possible get images taken with the same focal length as possible. Take them with image modeler in mind and ensure that you have as many 'basic defining cube form' points visible in 2-3 images per verticle corner. Make sure you have a clear view of the points that you know are correct to create the xyz rays of a coordinate system.

 

Start matching points that define the basic connected 'faces' of the cube form. Keep adding points after, if needed, to get a calibration. Set the XYZ coordinate system starting at the origin then out in the appropriate direction as IM calls out to pick the next point for that ray.

 

Create more Matching points as/if needed.

Define other 90 degree angles to tighten things up.

 

Refine the tolerance of match points....

 

Create the rest of the faces to be IM'd.

 

Export to 3D program.

 

In 3D program you have 2 options to make things better. Use camera mapping of the images and adjust the geometry to "square and Plumb" within the projection/s and tweek the camera a bit.

or Use Baked UV materials and tweek the geometry to make it right.

 

The first allows for a quicker-looser match in IM and addition of more geometry if your LOD requires it. Stuff like cornices. The second requires a good projection map fit in IM but allows you to move the geometry around without concern to how the cameras are placed in the 3D application.

 

I started with a simple box and worked up to more complex objects. And even then I tried one of a house last year, images taken by someone else, that looked very do-able but.... it was a b!+[h and everyone who tried failed.

 

Post up a couple images if you want and possibly find some assistance.

 

WDA

[Devin Johnston]

Thanks for the tips William, I foolishly tried to start out modeling a rather complex building which ended in disaster. After that I decided to try to model a box like you suggested using 3 different shots, surprisingly it worked rather well. Then I moved on to a simple building which I got to work after about an hour of trying different things. I think it will be pretty useful for scenes where you don't have to get to close to the IM'ed object, or where you don't have to see the building from every direction.

[quizzy]

devin, dont know abou the exact amount of buildings. but the project took around 4 months (1 person) We have developed a few methods to speed the whole process up significally. It prolly would take 4 weeks now....

[hockley91]

I think image modeler is great! I do wish it had some better modeling parameters and functions, but it's great for existing conditions.

 

I modeled an entire college campus from aerial photos and site photos and it's turned out great! I've saved so much time through image modeler creating the existing site structures. Would have taken me 3-4 times as long doing it as I used to do.

 

I still get frustrated with the calibration sometimes. I try to stay at 10-15 calibration points and I usually pick points that are not always building specific. So I'll hit some site reference points and building points. Sometimes, I'll get so frustrated I'll just start all over and pick different calibration markers. It's quirky and sometimes it works and sometimes it's a pain.

 

But, I've been using it since mid December and it's been a phenomenal asset to have and use.

[Chad Warner]

Adam-

 

can you post some examples of your end product?