SonotaCo.JP
SonotaCo Network Japan Forum
SonotaCo.JP Forum Index
homeTop Page  FAQFAQ   検索検索   メンバーリストメンバーリスト   ユーザーグループユーザーグループ   登録する登録する 
 プロフィールプロフィール   プライベートメッセージをチェックするプライベートメッセージをチェックする   ログインログイン 

not saving video

 
新しいトピックを投稿   トピックに返信    SonotaCo.JP Forum Index -> UFOCapture BBS in English
前のトピックを表示 :: 次のトピックを表示  
投稿者 メッセージ
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Thu Aug 07, 2008 7:37 am    記事の件名: not saving video 引用付きで返信

I'm catching sprites tonight but I just noticed that after a while no video (AVI) is written when it detects something. Running v.2.18. I had this problem before, a year ago, on a less powerful machine. I do not run out of disk space. It occurred tonight after 107 videos. I do use the windows trashbin which gets quite full (gigabytes). The problem is solved for a while when erasing some videos. Did anyone experience the same?

Oscar

_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Thu Aug 07, 2008 8:15 am    記事の件名: 引用付きで返信

ahh... it may be related to the slow object mask? I had it set to On, not using P/s, size 15. If I moved the camera it would perhaps define stars as slow objects.
Now I have P/s at 20. I used P/s before but often it would not mask objects especially pulsing airplanes.

Oscar

_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Fri Aug 08, 2008 1:20 pm    記事の件名: 引用付きで返信

Hi Oscar
Quote:
I just noticed that after a while no video (AVI) is written when it detects something.

Yes, this may happen when the "slow object mask" cancels the clip.
In case of it, all clip files such as *P.bmp, *M.bmp must be deleted at the same time.
If you cannot find AVI file only, it is a new mystery.
Quote:
it would not mask objects especially pulsing airplanes.

Yes, slow object mask is not perfect.
If the flushing interval is too long, it will not be masked(This could not be tuned).
Moving speed should be adjusted by P/s (pixel per second).
Also, if the flushing size is over the mask size, it will be detected. In this case, set SOsize appropreately.

I am writing this in the sea side resort house of my friend.
The network condition is very bad here. My carry phone can connect only a few minutes in an hour.
Also I forgot installing spell check program to my laptop Wink .
But the night sky is very dark Shocked , Gamma was useful for the first time.
トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Sat Aug 09, 2008 1:01 am    記事の件名: 引用付きで返信

Hi Sonotaco,

Good to hear you are enjoying the dark skies at the sea! Smile The world is too light-polluted nowadays and for no good reasons... Mad

Yes, not only the AVI but also no images were stored. The harddisk does nothing after such detection. The slow object mask seems tricky, its sensitivity apparently also depends on the detect level setting, which I had set 3 levels below what you can get if automatic. It works well for events, but not for the mask because noise can become marked as slow object. You can see this when enabling the checkbox before P/s after which the square boxes appear, they drift around. Of course, without slow object mask it detects all the time at such sensitive settings... many pixels surpass the detect level threshold. So with the slow object mask at large size this could lead to masking a large part of the image! On the other hand one could potentially use slow object mask for the most noisy pixels if you set the size very small (have to test it, of course it cannot be used for large objects then).

Usually in my case a new airplane triggers and data is saved... probably because it flashes and is dark in between. This likely depends on detect level setting because one needs to detect the airplane also during its 'dark' period between flashes. So, setting Detect level too high fails to detect this, and setting it too low makes too many apparent slow objects (noise).

But there is a condition that says not to store the video and images if, apparently, the new object (like an airplane coming into view) appears to be slow after its initial trigger release. This is a potentially "dangerous" situation as we could miss an event occurring during slow objects. True? Or does it still save data when other pixels elsewhere trigger in the frame (my "hand-trigger" test suggests not!) What is the criterion for not saving the data? Is there a relation to the tail setting? (decision for saving data must come before end of video?)

So, if the slow object mask sometimes takes an object, noise, or stars after I move the camera, it may not save the data (is this really necessary?)
The noise issue can be solved by not pushing detect level too far back (auto is fine, at some cost of detection sensitivity) but the star issue could be solved by requiring a slow object to move at least a certain amount of pixels (so besides maximum P/s also a minimum P/s - they should not stand still). Currently the only way is to stop the detection and restart it, and the condition can go unnoticed if the Slow Object Mask is used without the checkbox after it, as the squares are not visible.

Excuse me for asking all those technical details, but I just want to understand in order to get the best detection sensitivity, limiting the possible side effects.

Oscar

[edited: found some answers in the user manual]

_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Sun Aug 10, 2008 8:41 am    記事の件名: 引用付きで返信

Hi Oscar
Thank you for deep questions. I am back here in Tokyo.
I also hoping better detection sensitivity always, but stability of long tern detection
is also very important. UFOCapture should have balance of them under the limitation of
current environment especially camera sensitivity and processor performance.

UFOCapture is made to capture everything but known objects or phenomena that is not
welcome and the characteristic is well known.
UFOCapture does not recognize the target object(meteor or sprite) but recognize unwanted
objects and eliminate them.
We cannot recognize unknown objects, and the automated recognition never be perfect,
therefore, it is safety and may be the only way to capture unknown objects.

Slow object mask was made to ignore air planes or satellites that moves slower than
usual meteors but faster than fixed star, is not so big size, and continues long time.
Before the implementation of this function, we simply record all events or simply abandon
long events. There we learned the characteristic of air planes.
They are not so many, the simultaneous occurrence with meteors is rather rare,
but they remain very long time (even in minutes) in FOV and spend a lot of HDD space.
That was the big pragmatical problem. Therefore partial masking of these object was made
as an essential function.

But this function is not suit for noise at all.
Though it seems effective for noise, by some special settings, the characteristic(ex frequency)
is quite different from that of air planes.
It is natural that there may happen many troubles when you want to reduce noise event by this function.


Lightningwizard wrote:
its sensitivity apparently also depends on the detect level setting,
which I had set 3 levels below what you can get if automatic.

I think this is the cause of your question.
You are seeking inside the background noise. But it is out of current design.
Sporadic CCD noise recognition is not implemented yet.
I still recommend to fix size=2 or 3 and tune level appropriately
...means level = noise level + 2 or more.

Lightningwizard wrote:
On the other hand one could potentially use slow object mask for the most noisy pixels if you set the size very small (have to test it, of course it cannot be used for large objects then).

Yes, slow object mask has side effects. In the mode, hit point counting method is changed and isolated pixel noise will be ignored more often. But please do not count on it. The behavior is not planned and may have troubles.

Lightningwizard wrote:
one needs to detect the airplane also during its 'dark' period between flashes.

No, there is no problem if the dark period is completely invisible.
That is why slow object mask needs 3 seconds to decide its velocity.

Lightningwizard wrote:
But there is a condition that says not to store the video and images if, apparently, the new object (like an airplane coming into view) appears to be slow after its initial trigger release. This is a potentially "dangerous" situation as we could miss an event occurring during slow objects. True?

The cancellation of slow object mask happens only for the clips that was started by the slow object itself.
Therefore the clips that is triggered by other events or triggered while slow objects are flying will not be cancelled.
But precisely, there is short dangerous timing. That is the gray frame period of the first slow object.
It is the timing of an slow object appears (recording begins by it) but the velocity has not been measured yet.
Other events that happens within the period (less than 3 seconds) will be lost if the cancellation is happened.
I did not care about this because it happens very rare for airplanes.

There are many considerable improvements.
The recognition of pixel noise may be the direct approach to improve the detection ability.
The smoothing after taking difference that you are saying before may have similar effect.
Birds and insects recognition using the linearity is also pending problem.
Clouds recognition may be the most difficult matter.
The handling of High Definition video stream with tremendous data flow is also upcoming problem.
I do not know which is important and possible now.
トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Sun Aug 31, 2008 8:46 pm    記事の件名: 引用付きで返信

Last night I observed 39 sprites! My experiences:

One sprite happened as I watched the screen but it did not trigger! I did not remember if there were slow objects. Another period I heard triggers, hard disk working, when I looked in the directory there was no new movie. As there were hundreds of airplanes all night and movies of events are recorded during such period, I think maybe some part of the program hung. You can test it by waving your hand in front of the lens. I restarted the program.

The airplanes are quite a problem here in this area, especially during mid to late evening. At some point I had 8 of them in view. And many are hidden and give a pulse once in a while.
Despite slow object mask, the planes arriving into view trigger writing the files. Then, they should become blocked, but it is not always the case. After saving the movie data there often is a small interruption in the video stream. When it runs again, the mask stays behind and the slow object triggers during the new grey markers. I have not found a speed setting to counter this, maybe I should set it much higher, like 100. But I think the mask does not extrapolate once the video stream is interrupted, it uses the last position. The video is then saved again, etc. First I used uncompressed AVI where the bottleneck is the disk, now XviD with bottleneck CPU... but maybe only on Vista. I guess there is not much to do about it.
Since using XviD, I occasionally get a crash "Cannot write U->writer stream" and I have to restart the program. The latest video does appear to be written correctly. But for unattended capturing XviD may for this reason be a problem.

Oscar

_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
特定期間内の記事を表示:   
新しいトピックを投稿   トピックに返信    SonotaCo.JP Forum Index -> UFOCapture BBS in English All times are GMT + 9 Hours
Page 1 of 1

 
移動先:  
新規投稿: 不可
返信投稿: 不可
記事編集: 不可
記事削除: 不可
投票参加: 不可
このフォーラムで添付ファイルを投稿 できません
このフォーラムでファイルをダウンロード できます


Powered by phpBB © 2001, 2005 phpBB Group
Copyright ©2004 SonotaCo Network. All Rights Reserved.