For processing photographs into 3D models, this project uses the software Agisoft. Agisoft allows a high degree of control throughout the processing workflow, which is ideal to get data about the object and for its replication across objects. The reason that so many photos are taken is because the software uses pixel overlap in order to determine where a camera is facing. The below example shows where there was not enough overlap – the software recognized that there was a curve to camera circuits, however it did not understand that the object was rotated every photo.
This next example shows where the software correctly recognizes the rotation, there is much more overlap in these photos.
Throughout the processing, many photos are removed from the model. This is to make sure that the photos that do exist are very accurate and process correctly. In fact, around 90% of the points in the initial model will be deleted, leaving the most accurate projections.
Once these points are selected, the model is changed from a point cloud (A) to a dense cloud (B) to a solid (C), and finally to a textured model. During each step in this process, more and more points are refined or deleted.