A peek behind the scenes – how C3 did the 3D

Hello everyone! Hope you had a great Easter weekend!  As you head from Luca and Ludvig, the 3D maps beta was made possible by Nokia’s partnership with C3. Today we have some insights on the 3D imaging process that C3 used on the 3D maps project. C3 uses modern camera ...

Hello everyone! Hope you had a great Easter weekend!  As you head from Luca and Ludvig, the 3D maps beta was made possible by Nokia’s partnership with C3. Today we have some insights on the 3D imaging process that C3 used on the 3D maps project.

C3 uses modern camera equipment to capture as many as one image per second of the same object from up to 100 different angles. The images are then used to automatically reproduce the shape of the objects with very high accuracy. After that, our image processing software automatically drapes each shape with the texture chosen from the pictures of each object. Since they use the same process for all objects – buildings, houses, trees, and hills – the result is a seamless canvas of 3D-data where the resolution (8 to 12 centimeters per pixel) and quality is consistent over the entire model. This is the secret to C3 maps’ realistic look.

 

Click here to view the embedded video.

 

Traditional 3D models are made by hand using two components: terrain and building objects.The imagery that covers the terrain layer is only captured from above (not from the sides) and the resolution varies. Since the images are captured from high above, the colors of the images are usually very pale, appearing washed-out. The building layer, for example, is created by a human operator, who must select the shape of every roof of every building using a drawing or modeling tool.  They then try to drape the buildings with oblique imagery taken from four different angles, but with less frequency than C3. Then there is a final manual step where an operator must control and adjust the drape for each building side.

One common problem with this technique is that trees or other objects that are in front of a facade get stuck on the face of the facade. Also, since the shape of the building is simplified, the buildings tend to look cartoonish, or Lego-like.  Another problem is that since the imagery of the terrain is not captured on the same date and same flight at exactly the same height there is always a difference in visual impression between terrain and buildings. That also makes it more cartoonish and less realistic.

In contrast, with C3 the entire process is done by computers which delivers much more realistic models.

 

Click here to view the embedded video.

 

Only a single person is required to manually retouch the shapes of 10 to 40 objects in an average city, usually because they are important landmarks, compared to the traditional approach, which requires thousands of people to generate a 3D model.  As a result C3’s end product is uniform and photo-realistic.  To make the maps look good from the street perspective, in 3D maps we take the user to the street level view – a virtual walk on the park, if you will.

Source : http://blog.ovi.com/2011/04/28/a-peek-behind-the-s...