stats and diagnostics questions
Posted: Mon Dec 13, 2004 2:57 pm
I'm having a little problem setting the threshold values to get the desired alarm results. I'm looking at cars passing on a narrow street, so we aren't talking about subtile effects. In many cases, the alarm won't go off until the moving object is well into the frame. In addition, alarms seem to be more often triggered, especially for black objects (cars), by the re-exposure of the background image after the object has passed.
To help figure out what is going on, I turned on the diagnostic images and started looking through them.
I had a question about the diagnostic images, their meaning and location.
(I currently have a 20 frame lead-in set)
These image files show up in the events/<monitor>/<event> directories
0021-capture.jpg - the current image
0021-diag-r.jpg.jpg - the reference image which is the sum of the previous x images?
0021-diag-d.jpg.jpg - the difference frame between the current image and reference image
0021-diag-1-1.jpg.jpg - I'm guessing this exists but can't find it
0021-diag-1-2.jpg.jpg - the thresholded and filtered pixels
0021-diag-1-3.jpg.jpg - the identified blobs(?) (this seems to look very much like -2 above
0021-diag-1-4.jpg.jpg - thresholded blobs(?)
0021-analyse.jpg - the current image with the edge of the thresholded blobs drawn on it
Notes:
1. Not all the images are written all the time. The "analyse" frame is only written when there are statistics to report. (after the alarm frame count). It seems like there is nothing to "report" i.e. the diagnostic image is blank, it is not written.
2. I guess the files are written somewhere and then moved to the event diredtories. the ".jpg.jpg" extension I believe is a little bug and a result of this move (rename?).
3. The statistice seem only to be generated when an "analyse" frame is produced.
Can you correct my assumptions above?
Is there a little code hacking, (by me), that can be done to turn on diagnostic images and statistics prior to the first alarm frame so I can figure out why it is alarming so late?
Thanks
To help figure out what is going on, I turned on the diagnostic images and started looking through them.
I had a question about the diagnostic images, their meaning and location.
(I currently have a 20 frame lead-in set)
These image files show up in the events/<monitor>/<event> directories
0021-capture.jpg - the current image
0021-diag-r.jpg.jpg - the reference image which is the sum of the previous x images?
0021-diag-d.jpg.jpg - the difference frame between the current image and reference image
0021-diag-1-1.jpg.jpg - I'm guessing this exists but can't find it
0021-diag-1-2.jpg.jpg - the thresholded and filtered pixels
0021-diag-1-3.jpg.jpg - the identified blobs(?) (this seems to look very much like -2 above
0021-diag-1-4.jpg.jpg - thresholded blobs(?)
0021-analyse.jpg - the current image with the edge of the thresholded blobs drawn on it
Notes:
1. Not all the images are written all the time. The "analyse" frame is only written when there are statistics to report. (after the alarm frame count). It seems like there is nothing to "report" i.e. the diagnostic image is blank, it is not written.
2. I guess the files are written somewhere and then moved to the event diredtories. the ".jpg.jpg" extension I believe is a little bug and a result of this move (rename?).
3. The statistice seem only to be generated when an "analyse" frame is produced.
Can you correct my assumptions above?
Is there a little code hacking, (by me), that can be done to turn on diagnostic images and statistics prior to the first alarm frame so I can figure out why it is alarming so late?
Thanks