Skip to content

Help and Advice in Setting Up Movement Detection

Hi. Any help greatly appreciated. Love the software but i just cant get it to capture movement for all the area on one of my cameras. To set the scene the image is made up of 2 thirds from the bottom is garden then there is a fence and then tarmac with garage and cars in the top third of the image. When i set it to capture movement it seems fine with the bottom 2 thirds if anything moves but if you walk up to the car part of the garden it seems to stop detection in the car area. This includes the cars actually moving which seems fundamental to me. I have increased the sensitivity and masked out some trees to avoid them causing any movement but this still seems to let all hell break loose and i end up with many videos for minor movements to compensate to get the car movement detected? Am i missing something? I have got friends / family to walk up the garden and when the sensitivity is at 50 sure enough as soon as the person enters the car area it stops detection. Out of interest how does the camera detect movement as i realise that the movement at the top of the camera is smaller as it is further away but it seems wrong to have to increase the sensitivity especially for a car moving as an example? Thanks again.

Comments

  • Hi, the closer the moving object is, the bigger it is in the camera frame and therefore the more motion it will cause, triggering SecuritySpy. It sounds like the area you are referring to is too far away, and therefore the amount of motion generated by a person or car moving there is too small in the frame for it to be detected. SecuritySpy ignores such small movements as they are generally not interesting events. If you would like us to look into this further, please email us and include the camera's image (right-click on the camera in SecuritySpy and select "Save as JPEG"), and if a captured movie file is small enough to send by email please send one too.
  • I'm in general very happy with SS but I'm having a similar problem. After a LOT of fiddling the false positives in the day are close to zero, but unfortunately at night it's not triggering at all. If I put it up to 70 sensitivity then it works when someone walks close by, but it also gets triggered by insects. It's a bit frustrating and I'm not really sure what else to try...
  • one thing I'd check is the onboard detection grid. My cameras, Avtech have their own detection grid and some of it is enabled by default. If you then make a detection mask on SS, then it's possible ( an expert might tell you otherwise ) that you have two detection grids competing with each other. I think this caused me a lot of problems when setting up my motion alerts. Also I found that people walking towards the camera took longer to register that people walking across the camera view. And again, with PIR sensors they react to actual 'movement' whereas cameras I believe react to pixel changes and any number of things can complicate pixel changes
  • All up and running with the registered 4 camera software. All seems to be going well and trying various experiments with the original issue i had and will update when i know more. The main problem is spiders funnily enough now. The Compro IP400P camera i bought has poor night vision. The Foscam FI8905E though has amazing night vision although was a little problematic with the POE.
  • @peter - I don't really understand how an onboard mask can clash with SS (unless you're uploading images and video to the same place) but in any case it's completely disabled on the cameras. Unfortunately it is the pixel changes which is the problem. It seems like it should be possible to do something a bit more sophisticated than count the number of changed pixels, but for the moment I'm investigating separate PIR (or other) motion sensors.
  • SecuritySpy's motion detection is more sophisticated than simply counting pixels. It does several things:

    - Counts changes in pixels between the current frame and previous one
    - Automatically adjusts sensitivity based on measurements of the amount of noise in the video stream, which is constantly updated
    - If enabled ("trigger time" setting), monitors these pixel changes over 1 or 2 seconds, and uses all the information gathered during this time period to decide whether to trigger.

    In most circumstances this gives good results, with a low false-negative and low false-positive rate. However it is designed to err on the side of more false-positives, in order not to miss out on any important motion is should be capturing. This will especially be the case if the camera is situated outside pointing at areas of moving plants/shadows etc. Make sure the camera is angled downwards (so as not to capture trees/sunlight/clouds), and use the mask feature to ignore any problem areas.

    Hope this helps!
  • Thanks a lot Ben - sorry for underestimating SecuritySpy.

    One thing we have is that the cameras record lots every dawn. My impression is the camera is adjusting the exposure back and forth a bit before it stabilises. This leads to a lot of video being recorded. For the foliage and shadows, we've done what we can, but it's still a pretty massive problem. Either we get ~50 videos per 24 hours, or we have it set so that it doesn't detect anyone at night.

    Hence the need to get a separate motion sensor (though I've not found anything I like the look of yet!)
  • Hi Ed, I can certainly appreciate the problem you are experiencing. There are two problems here that we've seen before: one is that when the lighting changes, some cameras adjust their exposure in such a way that the lightness of the image changes rapidly, triggering SecuritySpy's motion-detection. The other is that sensitivity settings that work well at night don't work so well in the day, due to the very different image characteristics provided by the camera in these different lighting situations.

    You can mitigate the second issue to some degree by installing better lighting around the camera at night. We are also working on this issue from our end by looking at automatic sunrise/sunset detection and auto-adjusting the motion detection parameters based on that.
  • Hi Ben,

    That would be great!

    Cheers,

    Ed
  • Hi, Following on from the above comments all still working well but as Ben mentions i am getting loads of emails relating to changes in light and shadow. As an example today i am running at 500 emails in relation to effectively the sun coming in and out. If it could be the camera then any work in that area would be great. Not the end of the world and still working great apart from this issue in the main. Lessened my original problem of movement further away by tweaking the settings and the camera and making sure an external light is activated as well.
  • The light and dark issue appears to be camera related as i have 2 different cameras one with numerous movements detected from what appears to be the change in light and the other camera only appears to be picking up actual movements.
  • Ratio is about 15:1 in terms of amount of false email alerts i get in relation to one camera compared with the other. In a 24 hour period one camera is on 100 movements and the other nearly 1600. Is there a camera tweak i could perhaps do? The camera not playing the game is a Foscam FI8905E. Thanks.
  • Hi Ben,

    I was wondering if you could allow for a feature to permit a 3 or even 4 second trigger time. I have been watching the motion activity and in most cases 3 seconds would cover the light changes with sun / cloud / passing cars e.c.t
  • Hi - yes it would be easy to add a 3-second or 4-second trigger time, I will put it on the "to do" list for the next update. The problem is that this feature uses a lot of memory (because video footage of duration equal to the trigger time has to be retained in a buffer in memory), which is why it's currently limited to 2s, however this isn't going to be a problem if there aren't too many cameras or the resolution or frame rate is not too high.
  • How about a Time-Of-Day decreased sensitivity setting? Brackets of time where the trigger time was increased. That would reduce the RAM overhead to just those times. You could even do sunrise/sunset calculations...
Sign In or Register to comment.