Where the ends were within about 30cm of each other I made them horizontal.
The only exception was the Temple of the Sun, which I drew with the sloping walls. They remain within 20-30cm of correct both at the top and the bottom. The walls of all buildings and terraces slope inwards slightly. I did allow myself one concession to simplify construction. I have attached two photos of the exact same 3 terraces on Wayna Picchu that demonstrate it. Sometimes it had me very confused until I realised the photos were of the same thing. One interesting thing my photo searching showed up was the amount of restoration that has been done. You will notice that the cave is also modelled.) (TO get to the top of Wayna Picchu you have to crawl through a narrow cave. For the parts of the mountain itself that were in the scan I drew numerous vertical lines of the correct height then connected up all the top tips of the lines into triangles. That allowed me to do many of the structures and terraces although again it was missing a lot of detail. I then manually drew connecting lines and pulled them into shape. I therefore imported photos from various angles, positioned them in the correct place and traced around the mountain outline. For some of them, such as the remote lower terraces, that was quite a task.įor Wayna Picchu (Huayna Picchu) I could not find any good contour maps. For those I had to scour the web for photos - mainly, searching for photos that showed the particular area.
Places where the laser, positioned up to half a km away, simply could not see. The big problem was that there were numerous structures, terraces, walkways, staircases etc that were missing from the scan. I then drew all the structures on the flat plan view in Sketchup and then used the Pull tool to pull them up to the correct height. The program also allows you to measure individual points so I printed off dozens of screen shots of the plan view then measured individually every significant point on the entire site and wrote the measurements on the plans. I took multiple plan view screen shots then imported them one at a time into Sketchup as images and positioned and sized them accurately.
In the end I used a program (Rapidform Explorer) that could produce a visualisation of the point cloud. There were so many lines and vertices nearby that it was impossible to draw a line where you wanted, hence my recent request for an enhancement to allow ALT or something to be pressed to override inference snapping. Trying to draw flat surfaces on the imported wavy surfaces was virtually impossible, partly because of inference snapping. That doesn't include Wayna Picchu, which is a separate model.
STEREOSCOPIC 3D MODE SHADWEN PLUS
The model came in at almost exactly that, plus another 600k for the high-res terrain texture. I set myself a goal of 5Mb as a maximum file size. Had I used it the model would have been something like half a Gb in size. The resultant data, even when simplified, was massive. It worked fine but it proved to be a pointless exercise. To start with I tried cutting the data into hundreds of pieces and triangulating and importing it.
There is a cloud.rb plugin that can import the point cloud as construction points but it causes Sketchup to grind to a halt with just a few thousand points. I expected to be able to just triangulate the points and import them, but it proved impossible due to the size. The actual files I used were stripped down ones with about 20M points in total. I used a laser scan of the site done by the University of Arkansas which produced a point cloud file of 80M points. It was a much bigger exercise than I anticipated. Karl, I have been working on the model for 18 months. Once the textures are down-sized by Google it would be very difficult to tell one from the other. What I have used is an actual photo of part of one wall.Īpart from a few buildings, in particular those that I did texture differently, the stonework is remarkably similar. I have photos of most of the structures and started using them as textures, but it soon became apparent that the resultant model size would be enormous.
0 kommentar(er)
