imagine you have correct bluperints with all 4 viewvs (f, r, t, s). (they are BW - not greyscale, and they’re the same resolution.) you need to model the thing. you run the script and - trada! you got the model.
it is just an idea (probably very hard to do), but it could work like this:
the script iterates through each pixel and when it finds a black one (for example in the top view), it searches in the same line for another black pixel to pair it. when it finds, script iterates through the last view, needed to define the point’s exact location. when it’s done, creates a vertex.
by this way it creates millions of vertices. someone else could create an algorithm for creating faces
why am i telling this to you? i don’t know… i would just ask someone who can write in python for blender and try it out. at first, you could use one image as a blueprint for all views (eg. image 32x32 px, only few black pixels)…
they have scanners just for autocad, but it is very far from automatic.
Not all lines in blueprints mean something in an actual plan, some of the lines are to help to define measurements.others are for showing the swing of doors, and the radius of things that turn.
I am not saying it is impossible, but to totally automate the process would be a very complex AI.
Perhaps before you scan, if you go around and black out the unneeded lines manually before scanning then entering the dimensions as it registers each line.
very interesting idea, I hope someone can think of a solution for this.
That is similar to how motion capture works. The difficulty is each pixel has at least 3 axis(xyz). So you would need to do this manually which is really just modeling with a blueprint background. There are cool programs that do photo metric modeling. Even saw a cool program called video trace (they are looking to go commerical) that used video to model.
