Reflections on a past field season

This past December I had a short field season in Arequipa Peru where I finished collecting data for my master’s thesis. Even though this season was short (only 2 weeks), I feel like I learned as much as I did during my summer field season (2 months). One of the major things I learned about was that creating 3D models using Agisoft is incredibly time consuming and requires an extreme amount of patience. One of the major problems that I was having is that the models would not render and would not build a correct sparse could which is the first step in creating a model. Previously I thought that the main cause to all my problems was a lack of light but during this season I figured out that the direction of the light is as important, if not more important. I went back through photos I have previously taken that were used to create models successfully and I began to notice a trend. The trend was that the background was not only black but was always out of focus. I compared the successful photos with the ones I had taken recently and found that in the photos that had not successfully built a sparse cloud the background was semi-visible. So after learning this information, I change some settings on my camera and make sure that the background is completely obscured by shadow. Success! I finally get my first sparse cloud to successfully render after countless hours and many failures.

Although I am not an expert on the inner-workings of Agisoft, I have spent some time using the program and have a hypothesis to why having the background completely obscured is important. The simple answer is that the methodology I employ is predicated on having a fixed camera and the object being modeled being placed on a turn table which allows for different angles to be captured. Hypothetically the only objects that should be moving throughout the sets of pictures will be the projectile point and the surface of the turn table on which the point rests. If the background doesn’t have any recognizable features and is just a black blob, the program will only focus on the moving parts within each picture. If the background is illuminated at all, the program will attempt to add the black background into the model and create a spare cloud that looks nothing like the original object.

Overall, I am excited to finally start my thesis writing and work and I look forward to posting updates on this blog.


Comments

Leave a Reply

Your email address will not be published. Required fields are marked *