The process of using Agisoft Photoscan proved to be a reasonably difficult feat. Overall, I have completed a total of six different models. All of which had problems and issues to overcome.
The general workflow, after downloading the pictures, will now be discussed. After capturing around one hundred photographs I opened each to critique if any editing was needed before importing into Photoscan. The photos in terms of brightness and shadows seemed to be fine, however a few photographs had problems with focus. A mistake I would not be able to correct post processing. An important lesson was learned at this point, to ensure that photos are in focus during the capturing stage. Although we had the remote control of the Canon using Shane’s laptop, I had not ensured continual quality in a small amount of the photos.
The post processing complete, which became a time for reflection rather than editing, I transferred the images into Photoscan. Another problem occurred as I could not use RAW images, only JPEG’s were importing to the software. At this stage I had confidence that the model would still remain photorealistic. The photos are now placed in the gridview below the main Photoscan panel, as seen below.
The first time creating the model I focused on using the magic wand as a masking tool. Opening each image into the main window, a single click allowed the magic wand to remove any obscure parts of the image, in this case the white area surrounding the pot. The image below shows the area designated by the magic wand. Although it seems to cover most of the image problems still arose.
A severe problem was found here in the form of an assumption. The assumption that the magic wand would suffice for the production of the model. It proved false, as I will show in an image below. The assumption wasted valuable time, the mesh cloud and points take time to complete. Further work in masking parts of the image had begun. Use was made of the rectangular selection tool and intelligent scissors. This allowed for a more precise look at the model. The image below illustrates the rectangular section in use.
This tool proves useful when there are large chunks which need to be deleted from an image. However, when a more precise deletion is needed the intelligent scissors prove to be the best tool. This allows for a user to select points which map around the object, ensuring that the first and the last point match, as seen below.
After all these steps had been completed, the workflow was completed. Finishing the dense cloud and mesh was part of this process. My laptop power only allowed for medium settings to be completed at this point. But, the outcome would not change hugely. Importantly, an error occurred at this point with missing parts of the bowl and essentially some floating sections found outside the main shape. Although the sections on the outside of the model could easily be deleted using the Photoscan cutting tool, it could not fix the missing pot sections.
I attempted to rectify this by finding the images which had not been completed correctly. A simple zoom out function allowed for this to be possible. However after many attempts I concluded that the images used was the main problem.
The texture selections found within Photoscan are interesting, allowing for greater detail and different viewings. However, the generic setting is perhaps the easiest to complete and looks suitable.
The final product of the archaeological pot was a disappoint with regard to the original thoughts of the project. The conclusion to be found here was to be more precise during the early stages of the project. Mistakes had been made with the capturing of the images which could not be improved at a later stage.
The use of Autodesk 123d Catch proved to be an easier process. The software only allows for 70 images to be used. These images were automatically uploaded and rendered using the software. The software does allow for a series of post processing once the model has been completed, these include squares and line selects. These allow for deletion and further processing. The mesh can be changed after the original one has been completed. The choices range from mobile, standard and maximum depending on the power of a computer. Video functions are also available with uploads to youtube and the possibility to save to desktop.It has to be said that the user interface of this software is a lot more accessible than Agisoft Photoscan. The primary screen is shown below.
The result of the model can be seen below. Although the majority of the model can be seen and there are no holes present, like the first model there are still remaining problems. It shows that the inside of the pot was not captured properly, meaning there is a hole on the inside of the pot. Yet again the software used to complete photogrammetry has proven that mistakes were made during the capturing process which could not be rectified at a later date.