In our line of work we often do 3D modeling of an object for visualization purposes. This approach is not always practical or precise. If the object is relatively simple and can be measured with caliper and other measuring instruments, traditional modeling gives wonderful results. But when the object is relatively complex and has organic or irregular shapes, traditional modeling takes a long time and gives poor quality results. That's why we've researched the different 3D scanning methods. It turned out that there are two options.
The first option was a purchase of a 3D scanner. This type of device is mainly divided into two groups, first suitable for scanning relatively compact objects, and second suitable for mapping and geological purposes. Both types of devices have the same flaws, they can not generate high-resolution texture for the 3D model and have relatively high cost.
The second option is the use of a digital camera and a technique called 3D Photogrammetry preferred by many film studios in Hollywood and around the world. To use this method of 3D scanning you need a camera, freeware software and powerful enough computer. This meant that we could experiment with it without investment. However we must emphasize that the 3D Photogrammetry method has two disadvantages to be taken into account. The first is that the scanned object should be captured from as many angles as possible, and during this time everything in the frames should be static. The second drawback is that transparent or highly reflective objects can not be scanned with it.
After clarifying the options we chose 3D Photogrammetry. We started with scans of small and simple objects. The results were good, even the time it took for the computer to generate a model from the photos was very acceptable. But these scans were only from 30 to 60 photos witch did not give us a clue what can expect from the process when the scans get bigger and have more than 1000 photos. For this reason, we have decided to make a 3D model of a relatively large object with as many photos as possible, just to get to see how far we can push the process. We wanted the object we would scan to be well known in the Kardjali district and we decided on “Thracian sanctuary Tatul”(Tatul). For shooting, we decided to use a drone (DJI Mavic Pro) that we have because we can easily shoot at different altitudes and also allows a much faster capture than a DSLR camera, which is critical. Then we decided we had to plan the shoot so we could get the best possible conditions that included. First, the weather had to be overcast, so there would be no hard shadows that moves and confuse the software. Second, we must choose the time when there are no people on the site who will also move and confuse the software. After a little more than a week of waiting on the right day, one Sunday we went to Tatul at 15:30. The plan worked perfectly there was no one else on the site, so we got to work. We managed to make three drone flights and shoot Tatul from many angels as possible and it took less than an hour. The only circumstance that could impact the quality of the scan was the fact that the clouds were not very dense, and occasionally the illumination intensified noticeably.
After reviewing the photos, it turned out that we have 1066 usable frames from the site. We went to the process of making the model. We have experimented how the various settings affect the quality and time for processing of the model, and how much RAM each operation requires. After almost two months of experimentation, the behavior of the software and the results became predictable for us. Than we took the most detailed model that we were able to generate and simplified it from 40 million polygons to 3 million polygons, because the additional polygons did not bring any detail, and is difficult to work with such a complex object. We generated a texture for the 3D model and made the final animation.
- 3D Photogrammetry can deliver high quality model.
- The software is very well optimized and allows you to scan large objects with relatively little memory but requires a lot of processing time.
- This scanning technique is the best for such application and is not accident that many studios in the film industry prefer it.