People keeping popping in and out - please help!

[SyRenity]

Hi.

I'm trying to fix this for the past month, but still to no avail. The people in my recordings are keeping to constantly pop-in and out in the middle of the recordings. They appear out of no where, and disappear as well.

My settings for the camera as follows (tuned during this month):

Reference image Blend %ge - 50
Warmup Frames - 25
Image buffer size - 70 frames
Pre Event Image Buffer - 69
Post Event Image Buffer - 60
Alarm Frame Count - 1


Zone settings:
Alarm Check Method - Blobs
Min/Max Pixel Threshold (0-255) - 15, empty
Filter Width/Height (pixels) - 3, 3
Min/Max Alarmed Area - 2,0
Min/Max Filtered Area - 2,0
Min/Max Blob Area - 1, empty
Min/Max Blobs - 1, empty

I'll appreciate any help with this matter.

[jameswilson]

what framerate you running, and isnt that bleed very high?

[SyRenity]

Have a network camera with 15-20 fps. The blend % was set according to an entry in the wiki, which said the % could be set as high as 50.

I actually thought that % can help with fast moving objects.

[jameswilson]

afaik the blend being lower will cause better detection, its the level (i may be wrong here) where the latest image is 'blended' into the newer images to allow for chages in scene, ie light issues etc.

Also with a pre alarm of 69 and 20 fps thats only 3 and a bit seconds. Also 1 frame tolerance is very low. Id up your buffers to say 120 with an 80 pre alarm, drop the blend to 7 and try that mate

[haus]

What kind of network cam? I have a couple of IP cams that feed jpgs, and the frame rate has to be pretty low to work. I also see 2-3 second "freezes" now and then where the cam seems overloaded. I've noticed it all the time, including when motion starts. Do you have normal CCTV cams (with video out rather than IP cams) to test and see if you experience the same issue? I wonder if the problem could be isolated to the cam and not ZM. Just curious...

[kamand]

I also see 2-3 second "freezes" now and then where the cam seems overloaded. I've noticed it all the time, including when motion starts.

Another possible solution - tune mysql performance.
Regards.

[Flash_]

I'd think the same as haus, the camera is at fault, but could also be server overload.

Could also be an intermittent wiring fault - run ping to that camera for a few hours or even a day and see if you get any dropped packets. If you're running over a wireless link that also increases the chance of data loss.

What's the average server load? (Something like munin makes it easy to monitor over a period)

What's the make and model of the camera?

Is the network reliable? (ping test over a long period)

Is the network saturated with other traffic?

What's the pre-event buffer? (If 0, modect won't show anyone coming into range and an event will start from when motion was detected.)

What happens when you use of the tested preset motion sensing settings?

[jameswilson]

having 70 buffers and using 69 of them as prealarm is asking for dropped frames, as the system will at time be behind and need this buffer, if its behind and the meotion starts it wont have 69 pre alarm frames available. Increasing this buffer will help

[W.]

... and 50% blend %age means that more than half of frame has to change between two consecutive frames, which human being can do only if he/she hops into frame really fast and fills more than half the frame at the same time. go for 7-15%. (smaller blend %age means more sensitive detection)

[jameswilson]

I dont think its that mate, as thats in the zone settings bleed is to do with the new image merging into the old image, else ptz cameras couldnt be used for detection as the image would always be different. On a static cam a high bleed isnt as bad but means light changes / scene changes would cause alarms. But ive never used a bleed above 10% myself

[W.]

I dont think its that mate, as thats in the zone settings bleed is to do with the new image merging into the old image, else ptz cameras couldnt be used for detection as the image would always be different. On a static cam a high bleed isnt as bad but means light changes / scene changes would cause alarms. But ive never used a bleed above 10% myself

it definitely is not root cause and number of buffers vs number of preevent frames is more likely to be the one, but still it is somewhere on the cause list

[SyRenity]

Hi.

I used the blend %ge specified on this Wiki page:

http://www.zoneminder.com/wiki/index.ph ... por_Trails

"However, I have found that values greater than 10 work quite well. Indeed, I have used a value of 50, which shouldn't work, but is giving me acceptable results."


So, the advice is to have it back on the default 7%?


I will increase the buffers to 120, with 119 pre-event (according to Phil recommendation in another thread) and 120 post-event, and let you know how it goes.

About the used camera, it's the old DCS-900 by DLink (don't know if they still manufacture them). The connection is a LAN connection, no ping drops AFAIK.

[W.]

I will increase the buffers to 120, with 119 pre-event (according to Phil recommendation in another thread) and 120 post-event, and let you know how it goes.

and you will get exactly the same effect. leave buffers at 70 and decrease prealarm framecount to say 50, that might help you

[jameswilson]

you will get the same problem but the is more pre event frames for it to drop. If you raise the bufers to 120 then have say 90 prealarm frames it gives you tolerance.

[kamand]

SyRenity
http://www.day32.com/MySQL/tuning-primer.sh
get this script and post results for sections:
[MEMORY USAGE] and [KEY BUFFER]
here are my results:

Code: Select all

MySQL Version 4.1.12-log i586
Uptime = 1 days 23 hrs 45 min 5 sec
Avg. qps = 9
Total Questions = 1710605
Threads Connected = 12

SLOW QUERIES
Current long_query_time = 10 sec.
You have 0 out of 1710620 that take longer than 10 sec. to complete
The slow query log is enabled.
Your long_query_time may be too high, I typically set this under 5 sec.

MEMORY USAGE
Max Memory Ever Allocated : 234 M
Configured Max Per-thread Buffers : 403 M
Configured Max Global Buffers : 138 M
Configured Max Memory Limit : 541 M
Total System Memory : 1.49 G
Max memory limit seem to be within acceptable norms

KEY BUFFER
Current MyISAM index space = 457 M
Current key_buffer_size = 128 M
Key cache miss rate is 1 : 1837
Key buffer fill ratio = 100.00 %
You could increase key_buffer_size
It is safe to raise this up to 1/4 of total system memory;
assuming this is a dedicated database server.

Before tuning mysql I had this records in mysql logs (cause - zmaudit):

Code: Select all

# Time: 071115 12:08:23
# User@Host: zmuser[zmuser] @ localhost []
# Query_time: 23  Lock_time: 0  Rows_sent: 950  Rows_examined: 22416123
select distinct EventId from Frames left join Events on Frames.EventId = Events.Id where isnull(Events.Id) group by EventId;
# Time: 071115 12:39:23
# User@Host: zmuser[zmuser] @ localhost []
# Query_time: 23  Lock_time: 0  Rows_sent: 400  Rows_examined: 22434612
select distinct EventId from Frames left join Events on Frames.EventId = Events.Id where isnull(Events.Id) group by EventId;
# Time: 071115 13:10:55
# User@Host: zmuser[zmuser] @ localhost []
# Query_time: 24  Lock_time: 0  Rows_sent: 1200  Rows_examined: 22452066
select distinct EventId from Frames left join Events on Frames.EventId = Events.Id where isnull(Events.Id) group by EventId;
# Time: 071115 13:42:11
# User@Host: zmuser[zmuser] @ localhost []
# Query_time: 24  Lock_time: 0  Rows_sent: 750  Rows_examined: 22456198
select distinct EventId from Frames left join Events on Frames.EventId = Events.Id where isnull(Events.Id) group by EventId;
# Time: 071115 14:13:35
# User@Host: zmuser[zmuser] @ localhost []
# Query_time: 24  Lock_time: 0  Rows_sent: 950  Rows_examined: 22460926
select distinct EventId from Frames left join Events on Frames.EventId = Events.Id where isnull(Events.Id) group by EventId;
# Time: 071115 14:45:04
# User@Host: zmuser[zmuser] @ localhost []
# Query_time: 26  Lock_time: 0  Rows_sent: 600  Rows_examined: 22470471
select distinct EventId from Frames left join Events on Frames.EventId = Events.Id where isnull(Events.Id) group by EventId;
# Time: 071115 15:15:56
# User@Host: zmuser[zmuser] @ localhost []
# Query_time: 25  Lock_time: 0  Rows_sent: 0  Rows_examined: 22532786
select distinct EventId from Frames left join Events on Frames.EventId = Events.Id where isnull(Events.Id) group by EventId;

And sometimes at the same time I saw something like you (very infrequently).
Increasing key_buffer_size seems did a work for me. All my queries now under 5-6 seconds (from 20-25).
Regards.