3D model [ with textures ] - from google earth or similar.

Hi guys, i´m wondering if exist already any approach that allow us to extract 3d models with textures if possible in high definition high-poly [city buildings] from google earth or similar, using grasshopper & Rhino. I already explore Cadtoearth for rhino but definitely it's not what i'm looking for.

Thanks in Advance.

Replies to This Discussion

Permalink Reply by Kim hauer on July 13, 2016 at 8:53am

There is also this texture site: http://www.textures.com/category/buildings/1926

Permalink Reply by Joaquim Filipe Reis on July 13, 2016 at 3:25pm

Kim, are you talking serious !? :)

Permalink Reply by Kim hauer on July 13, 2016 at 5:26pm

Clearly you have a picture in your head of what your after, which you describe only in words. I have not seen a picture yet of what your after, do you see the dilemma ?

Permalink Reply by Joaquim Filipe Reis on July 14, 2016 at 6:18am

:) Ok, I need a precise location. The best result i reach so far with 3d photogrammetry. The problem is, once i´m using input images with low resolution, I will never obtain an high poly resolution, for that i probably I need some kind of drone associated.

Permalink Reply by ng5 Alex on July 14, 2016 at 6:35am

Seeing your previous posts i think this is you best chance. working with sfm point cloud. then either continue working with the point cloud (4d with color info) or with the resulting textured meshed model.

you need high rez photos, where many pixels describe as much detail as possible. meaning that a low resolution photo, will squeeze lets say a facade opening in a pixel, but since you need detail, need it to be captured in more pixels. Problem is that if the area you need is like the one in screenshot, geometry result is going to be very heavy. you will need a beefy pc to be able to work. also another shortcoming if you choose to work with a single mesh is that the texture will be huge. you need to check your cards max texture size support.

you will also need to form a capture strategy. Reality capture forum and facebook group have lots of info about such scenarios. chances are that you might need more than one run. but thats another discussion.

hope this helps.

best

alex

Permalink Reply by Joaquim Filipe Reis on July 14, 2016 at 6:44am

Hi Alex,

Thank you for your tips. This is exactly what i think about my issue. If have access to drones or laser captures in the place I would have no problems. So i need to find some other solution. At this time i´m working with a new 3d photogrammetry technology software [Capturing reallity] that operates really fast, for the previously image I take 131 images, it takes around 8 minutes to computer process the request, point clouds, texturing and mesh.

Permalink Reply by ng5 Alex on July 14, 2016 at 6:52am

what is the source of these photos? 131 images, seems not much for such area, but i would except a better result than the screenshot. you can visit Reality capture arena, facebook group. Milos Lukac will get you on track for the settings you need to get the best possible result from these images. Then see if you need more via a drone.

Laser capture is another story itself. You will to position you laser scanner in many places, highs and lows to be able to fully capture the area. Combine scans etc. It is more tested and accurate than Sfm, but Sfm is very flexible and pretty close to laser scanning accuracy.

Permalink Reply by Joaquim Filipe Reis on July 14, 2016 at 7:02am

I recorded one video from apple maps of the area that I intent to capture [3 minutes]. Then I use Motion 5 to calibrate the colors of the video [ color balance, hue-saturation, gamma, etc .. ] and I extract picture frames with one small app[adapter]. I get a total of 131 images, [ So in fact they are the same as screenshot] and I use this image as imagery to 3d photogrammetry. Probably this is incorrect, but it's how far I can go without drones in place.

Permalink Reply by Mohamed Naeim on July 13, 2016 at 9:33pm

What Kimm suggested is a kind of good step forward, if you are working in a big scheme

there is something called ESRI city engine,you can download a city 3d model for it format, and then use Kim website to download similar maps to apply to your buildings facades parametrically in this one

for example if you are working in the city of Paris, then you can download 7 or 8 different french facades maps and then apply it to the whole city or district

this is just another idea if you are not strict for specific building

https://www.youtube.com/watch?v=TucS6vLJVcY

I think its also possible to map images maps into masses in Grasshopper!!!

Permalink Reply by Joaquim Filipe Reis on July 14, 2016 at 6:27am

Mohamed,

I believe this approach is incorrect and it's in fact too much inaccurate. I already test all this stuff inclusive ESRI, Aero3dpro, and so many others, the result it´s not good at all, once I´m sure that the workflow it´s not the most indicated. On the comment above I post an image where I explain how far I gone until now. but with the technology that i´m using, it required more quality of the image or some kind of laser stuff for a precise representation. For that it requires a large budget, so i'm looking for other solutions.

Permalink Reply by ng5 Alex on July 14, 2016 at 7:10am

3 minutes of video yields only 131 images? sorry i am not familiar with the process you describe. But getting all possible frames from video the better. also carefully edit the video, make only color based changes, as any other change could mess with the reconstruction algorithm of RC.

Permalink Reply by Joaquim Filipe Reis on July 14, 2016 at 7:15am

It´s only an option as a draft test, but I can choose all the possible frames of the video. :)