Acquisiton time at classification and runtime problem

Prev Go to previous topic

Go to next topic

Last Post 30 Nov 2013 02:49 PM by anon

1 Replies

Sort:


You are not authorized to post a reply.

Author

Messages

anon

New Member

Posts:

30 Nov 2013 02:49 PM
Hi, I would like to ask some help in two cases because I'm clueless. In my College Thesis I would like to write about measuring hyper spectral image classification (all supervised) running times. The Acquisition time status in the .hdr file is always Unknown. The IDL function for the Acquisition Time is runs, but without any result (http://www.exelisvis.com/docs/ENVITime.html). The Activity Monitor only sees the X11 runtime resource usages. How could I measure the classification running times in merit way? The other question is: because of the speed factor, I'm using a strong Macintosh with 12 core and 48GB RAM, to classify the high-resolution images with Support Vector Machine supervised classifier. However, at the classification process, the ENVI is using only one (!) core from 12, and the runtime is unacceptable. What could be the reason? Is the application capable to take advantage of all the resources? How about multiprocessing? How can I speed up the classification process? I'm using ENVI 5.0. I would appreciate for any advice! Regards, Judit

Deleted User

Basic Member

Posts:228

11 Dec 2013 01:15 PM
The acquisition time in the ENVI header is intended to store the date and time that the image was collected by the sensor. Not the time that the image was generated by some kind of processing in ENVI. That is why the field is blank in your classification images. To determine the running time for classifications in ENVI, you may want to program some ENVI batch code that uses the IDL systime routine to record the time before and after running the classification. Many types of processing in ENVI will inherit the multithreading provided by the IDL routines used in their source code. We have a help article that describes multithreading in ENVI and IDL here: http://www.exelisvis.com/...cleID/3252/3252.aspx Not all IDL routines are multithreaded, so not everything that ENVI's SVM classifier does will be multithreaded. In particular, File I/O uses a single thread. Also, it is possible for the overhead required by multithreading to be more than the advantages of using multiple processors, and in this case, ENVI will decline to use multiple threads. If your dataset is small, this may be the case. - Peg Exelis VIS Tech Support

You are not authorized to post a reply.