Quote:The main open question refers to the time processing. Some of them are trying to speed up the run improving the HW but everything looks a bit empirical from my point of view.
Do you have any special suggestion in order to speed up the computing time? Are you taking into account the possibility to work on an improved version addressed to the operational customers?
Processing hyperspectral data with ATCOR is indeed time consuming and we did our best to increase the speed so far. Unfortunately, if processing in terrain, the speed is still too low if it comes to spatial high resolution data and we don't have any other strategy than what users are trying. My typical recommendation is:
- use 64bit systems with at least 16GB of RAM
- increase the memory allocation to IDL to a maximum (for windows)
- work on high performance Linux/MacOSX systems instead of Windows
- use SSD disks
- go for tiled processing if the size of a single band exceeds about 1 GB.
- set up a batch process for the processing such that it can be run over night.
We are thinking about an option to reduce the processing time by using the flat atcor code instead of rugged and feeding some more input layers - but this will still take some time to develop.
Search
Member List
Calendar
Help
