Pre-processing

1. Download images from SD card onto an SSD hard-drive

   1. Back up the data onto a secondary cloud drive (e.g. google.drive or Cyverse)!!

2. Use an established naming scheme for project consistency e.g.

   mkdir project_location\date_of_flight\flight_number

Setting up PhotoScan for Network processing

In order to run PhotoScan in network (cluster) mode each node must have a separate license of PhotoScan installed.

The setup is fairly straight forward. Ideally, the server + nodes are powerful machines (e.g. SNRE Snow and DarkTower).

Setup can be done in the GUI, but also needs to be run in a terminal window

c:
cd /program files/agisoft/photoscan pro/
./photoscan.exe --server --control 10.192.196.167 --dispatch 10.192.196.167

For the worker nodes:

c:
cd /program files/agisoft/photoscan pro/
./photoscan.exe --node --dispatch 10.192.196.167

Single project Processing

1. Open Photoscan
2. Load images from folder or individually
   
3. Align photos
   1. Select Low or Moderate Quality
   2. Click on the Advanced tab
      1. Select mild or disable depth filtering.
         1. Fair Warning: disabling depth filter will cause hundreds of thousands of floating outlier points

4. Build Dense Point Cloud

1. 1. Select a quality level
      1. Ultra High = native resolutoin of the images
      2. High = 4x lower resolution (2x2 pixel resampling), 2x lower precision
      3. Medium = 16x lower resolution (4x4 resampling), 4x lower precision
      4. Low = 64x lower resolution (8x8 resampling), 8x lower precision
      5. Lowest = 256x lower resolution (16x16 resampling), 16x lower precision

Note: This is the most computationally intensive portion of the job.

Side notes

I've tried to increase the number of SIFT points from 40,000 to 160,000 and the number of tie points to 32,000  in an attempt to reduce the amount of occlusion in vegetation during the dense point cloud generation. I was not sure of the results and the processing time was increased dramatically.