Taking a Photobatch

To take a photobatch and transmit it to the 3D Spatial team, you need the following equipment:

  • a Samsung Galaxy 2 (Android-based) camera, provided by the project;
  • a set of coded targets (CTs); and
  • an RTK-GNSS rover to record the coordinates of CTs.

After completing the excavation of the context and cleaning as much as possible, follow these steps to capture the photobatch:

  1. Lay out the CTs close to the boundary of and outside the context.
    1. CTs are said to be “coded” because they have machine-readable designs on them that are automatically detected in the software we use to create 3D models and their derivatives: Agisoft Metashape Pro. The coordinates of the CTs are measured with the GNSS, named according to a prefix + the CT number, and uploaded to Metashape. Metashape then identifies each CT in the photos and associates each CT with its uploaded coordinate data. This enables the software to georeference each model in real space as soon as it is initially processed.
    2. Use at least four CTs per context (three points are required for georeferencing, four provide some redundancy in case there is a problem with one of the targets). Most photos in a photobatch should capture at least one CT and preferably two CTs in each frame. Use enough coded targets to sufficiently surround the entire context. Again, the CTs should be close to the boundary of and outside the context—never inside it.
Image showing the ideal arrangement of coded targets around a context, from the original protocol by Brandon Olsen and Manny Moss.
  1. Request a prefix. Prefixes are letters of the alphabet assigned to each photobatch to distinguish it from other photobatches recorded the same day (the CTs are reused for multiple contexts throughout the day and CT numbers should not be duplicated within the day). Prefixes should be selected in communication with other area supervisors and the 3D Spatial team. Prefixes restart with A at the beginning of every day.
  2. The next two steps (measuring coordinates and taking photos) can be done in any order, but once the photos are taken DO NOT move the CTs until the model has been processed correctly.
    1. Measure CT coordinates: Use the GNSS to measure coordinates for each CT surrounding the context. The point of the GNSS survey pole should be seated in the center of each CT and the bubble level should be in the center circle. Two people make the work easier: one person monitors the bubble level and controls the GNSS; the other helps situate the point of the GNSS survey pole at the center of the CT and reads the CT number out loud. Each point is named using the prefix for the photobatch + the CT number (e.g., A84 or C128).
    2. Take overlapping photos: Begin taking oblique photographs of the context standing on the ground, moving in a left-to-right, then up-to-down (or vice versa) zig-zagging transect that covers the entirety of the context. Photographs must overlap significantly: a general rule of thumb is that 30–50% of the previous photograph should overlap the current photograph. Capture the entire context as well as all CTs such that the area covered is larger than the context itself. Most photos should contain at least one CT, and preferably more than one. Any vertical elements that might be associated with the spatial context should also be photographed vertically in the same pattern as the horizontal photographs (e.g., the top of a wall that the excavated deposit revealed once it was fully removed). Be thorough and careful; make sure your shadow and the shadows of other objects/people in the excavation area are not in the photographs, as this will obscure the clarity of the photobatch and possibly cause problems with the 3D model (see image below). When in doubt, take more rather than less photographs; however, also remember that the more photographs there are in the photobatch, the longer it will take to upload to the server and be processed. TIP: Bookending each photobatch with non-photobatch photos aids in selecting photobatches for upload and processing (e.g., a picture of a hand, foot, bucket, etc.).
As shown in this example, accidentally getting your shadow in photos can cause problems with the model and with the orthophoto down the line.
  1. Once all the coordinates and photographs are captured, upload the data to the server for processing.
    1. Photos can all be uploaded from the Samsung camera directly to a shared folder, if connected to the WiFi network. On the camera, select all the photos that belong to the photobatch and select Copy. Then, go into the folder for the server on the camera, and then into E:\gygaia\3dspatial\photobatches\processing. With the processing folder, create a new folder named according to the date and time of the photobatch. A photobatch completed on 2 July 2023 at 9:27 AM should go in a folder titled 202307020927; a photobatch taken the same day at 2:31 PM should go in a folder titled 202307021431. (NOTES: Do NOT round the time to the nearest 0 or 5 seconds; it is easier for the 3D Spatial team to distinguish photobatch numbers that are as unique as possible. Also, if you take a photobatch at the same time as another excavator, make sure to communicate clearly about who will use what filename to avoid confusion.) Once the photobatch folder is created, Paste the photos into it. 
    2. Coordinate data is automatically synced over the internet. Make sure the GNSS mobile phone is disconnected from the GNSS receiver and connected to the internet through either the Wi-Fi or cellular network. Once the data has been synced, it can be exported easily from the phone or on a computer using the browser-based Emlid Flow 360 app.
  2. Once the files begin uploading, contact the 3D Spatial team via Google Chat to let them know the following information:
    1. the context key associated with the photobatch;
    2. the prefix;
    3. the number of CTs used in the photobatch; and
    4. the number of photos in the photobatch.
  3. Once all data has been uploaded, the 3D Spatial team begins to process the photobatch. DO NOT move the CTs or begin excavating immediately adjacent to the photobatched context until the 3D Spatial team confirms that the model is good. Sometimes the 3D Spatial team will determine that additional photos are needed and will advise accordingly.
  4. If you take a photobatch for a discrete small object (vessel, find) or feature that requires both a “top” and a “bottom” model in quick succession, a few slight adjustments to the protocol include the following. 
    1. Make sure to wait for 3D Spatial team confirmation that the top photobatch is good before removing the object/feature.
    2. To save time, leave the CTs in place—undisturbed—when the object/feature is removed so that their coordinate data can be reused for the bottom photobatch without re-measuring them.
    3. As described elsewhere, save the photos and coordinates for the “top” and “bottom” models in their own subfolders, within the main photobatch folder (e.g., if the top photobatch is taken on July 2, 2023 at 9:27 AM, it will go in subfolder called “top” in 202307020927; regardless of when the bottom photobatch is taken (immediately or several days later), it will go in a subfolder called “bottom” in 202307020927).
    4. For very small contexts, miniature CTs are available. However, it is generally preferred that you use the regular sized CTs if possible because miniature CTs can be dislodged much more easily.

Powered by BetterDocs