Friday, July 26, 2013

Sara Holmqvist, Cléa Moulin & Freya Roe, students of our courses in Digital Archaeology (ARKN10-ARKN09), Lund University

Sara Holmqvist
Cléa Moulin
Freya Roe
Subject: Kärnan in Helsingborg.
Project goal: Create a 3D model of the entire outer surface of Kärnan.

Brief History:
   According to dendrochronological analysis, Kärnan was built during the 1310s. The tower is the only part of the one time expansive Danish castle that was used for three centuries. In stories from the Medieval period, Kärnan is said to be a seat of power in eastern Denmark, and thus experienced its fair share of battles and seiges.
   Kärnan got its present form as the result of a restoration that took place from 1893 to 1894. Having fallen out of use for at least a century, it was restored by the town- partially because of its dangerously ruined state. Before rebuilding, the tower was approx 30 meters high, compared to its current height of 35 m. Much like our model, the top part of the tower was completely reconstructed at this time. Thus, our model is of the present appearance of the tower, not a historical interpretation.

Our Process:
   During the first week of the assignment, all three of us traveled to Helsingborg to acquire pohotos of Kärnan. First, we took photos specifically of the bottom as the view is partially obstructed by a low wall surrounding most of the building. Next, we took pictures which included as much of the building as possible. Following this, we walked to the top of the building. Though we had our doubts about PhotoScan being able to create geometry for this part of Kärnan, we thought it would be best to at least try. We took the greatest amount of photos in this part of the acquisition.
Side tower being cleaned in MeshLab
   The second week was spent mostly trying to get a good result in PhotoScan. Even after attempting several different methods of processing the top, only one face of the side tower was useable. The bottom models and the models of the middle of the building were quite successful. We were able to begin aligning meshes of these two parts during this week as well.
   Naturally, the third week was spent doing everything else. By Wednesday we had reconstructed the top of Kärnan using 3ds Max and aligned it with our otherwise complete model. We began texturing the model on Thursday as well as setting the scene with VRay materials. At this point, we felt we could add to our model to make it feel like a more complete scene- such as stairs, the low surrounding wall, glass for all the windows, and grass.
Technical information:

When using Photoscan, we used all of the photos at first to align the cameras, but when building the geometry we sometimes had to disable half of them, otherwise there was a problem with the amount of space the process used.
All aligned meshes, imported to 3DS Max

    We aligned and cleaned the mesh mostly using MeshLab. However, there were not enough reference points to connect the bottom tower meshes with the rest of the tower, so we combined those in 3DS Max. Applying the Poisson filter was very problematic, and we ended up not using this filter at all. Texturing was not possible with MeshLab, so we ended up using only 3DS Max for that part of the project. Before importing into 3DS Max, we had to reduce the number of faces from 900,000 to 400,000.
   We used 3DS Max to reconstruct the majority of the top of Kärnan. Aside from having to scale much of the created components, this was very smooth. We also reconstructed the stairs and surroundings in this program. Texturing was a challenge because of all the different objects we used to create the model. We created IDs for each object that needed a specific texture, and applied VRay material bitmap to the MultiSub Object we created. 

Top, meshes only
Top, adding floor
Top, almost completely reconstructed

 Behold, our final rendering!

No comments:

Post a Comment