Autodesk Research
Blog

Some sample datasets
May 11, 2009

3D View

We are in the process of anonymizing the complete dataset, removing identifiable objects, license plates, etc. The full dataset is coming soon. In the interim, we are releasing a subset of the scans.

Landing

In this image, one can see a 2D representation of the 3D point cloud scan data collected in a single scan from one position. The scanner was placed at the bottom of the stairway to the 6th floor that appears prominently on the right-hand side of the image. The distortion in the image is similar to the distortion seen on a map of the Earth. For example, all the pixels along the bottom of the image form a small circle at the bottom of the 3D data set. When stretched out to form a 2D image, the small circle becomes a long line. This is similar to the way that Antarctica is shown as a long thin land mass at the bottom of a map, even though it is really a (mostly) circular land mass.

Office Space

Note that the brightness of a particular pixel does not represent the distance from the scanner but rather represents its reflectance to the laser. That is, in addition to generating a 3D value for each point captured, an associated brightness value is stored. On the images and scans shown here, we have made no changes or transformations to the data file we received from the scanner. In the center of this scan image, taken in the doorway of an office, the metal doorframe is quite reflective and shows a white "hot spot" in the middle where the laser was reflected significantly. However, if one examines the 3D data along the door frame, it is completely flat and contains the correct distance values.

Another noteworthy relationship is that objects closer to the scanner cast a larger "shadow". This means that, for example, the desk in the office hides quite a bit of the floor behind it. When looking at the 3D scan data from another angle, it becomes quite clear that no data is collected "behind" the desk and the resulting empty space in the data set looks like a shadow even though no lighting is present in the 3D point cloud data set. This is why it is helpful to have overlaps in multiple scans to better capture the surfaces that may be obscured from other angles.

Meeting Room

In the case of this meeting room, combining two scans, taken close to each entrance to the room, gives much better coverage of the room than could be achieved from a single scan in the center of the room. The Faro software can align multiple scans (as described in a previous article), so that a single point cloud data set can be created of complex spaces with multiple rooms and areas, from which geometric surfaces and other features can be created.

Lobby

Finally, the Autodesk lobby at 210 King is shown here also combined from multiple scans. Again, the use of multiple scans increases coverage around obstructions like the posts in the middle of the space and the corner wall in the middle of the L-shaped lobby.

The image content also contains some interesting artifacts. For example, at the top of the post near the chair, there is a flag that was swaying to one side from some internal air flow. However, the flag is distorted in a way similar to the distortion achieved when moving a document around on the scanning glass of a photocopier. This distortion is created by the different positions and shapes of the flag as it passes through the vertical scan trace at the moment that the scanner was pointing in that direction. These artifacts reveal the scanning technique and to some degree, the speed of the rotation of the scanner head.

Completing a scan of a building provides a great deal of data that could be used for many purposes. Our primary use is to provide a way to perform multitudes of measurements that could perhaps be done using traditional methods but would be incredibly time consuming and resource consuming. As the accuracy of our floor plans is unknown, we can now use the data set to properly measure, in all three dimensions, the size and shape of 210 King.

Download the full resolution scans below. To view the file, you will need to download Faro Scout LT viewer, a free download available from Faro.

Open a faro workspace file (.fws) in Faro Scout. Once loading the dataset(s) is complete, you can view the 3D point data by right-clicking on the folder that contains the scan files and selecting New View and then 3D View. To view the Planar projections as in the images above, you can right-click on the individual scans and select New View and then Planar View.

Navigation tools can be found in the toolbar at the top:

  • The Examine tool Examine allows you to orbit around the data.
  • The Pan 3D tool Pan 3D lets you move from side to side and up and down.
  • The scroll wheel lets you zoom in and out.

There are also preset views:

  • Canonical views along the primary axes can be access using the Top View, Right View, and Back View buttons Canonical Views.
  • A fit-to-view button Fit-To-View moves the camera far enough away so that all data points can be seen.

The right-click context menu in a 3D View provides more options under "Visibility Settings...", where the camera projection can be changed from perspective to orthographic, the background color can be changed, as well as the display size of the data points.

ZIP
Sample Scan: Fifth Floor Landing
ZIP - 61.195MB


ZIP
Sample Scan: Office Space
ZIP - 60.969MB


ZIP
Sample Scan: Meeting Room
ZIP - 91.343MB


ZIP
Sample Scan: Lobby
ZIP - 114.867MB


Previous | Next