Algorithms in motion detection


#1

Hi

In the program you have option to set Motion Algorithm to 6 different types, but in the manual you only describe 3. Can you elaborate on the other 3? I know they are somewhat self-describing, but a brief description and some examples for when they’re good would be appreciated!

Regards,
FSO


#2

This is part of what was added this week in the user manual (the update hasn’t been yet published).

The only mode that should really be of interest here is the blob detection. Instead of just measuring the amount of noise on the screen, it will look forward for big moving objects. You can right click on the camera and select View -> Motion View this will actually show you for each mode what it’s doing / detecting in a visual way. This helps understanding what it does and how to tune the settings based on your environment.

The person detection / face detection I suppose are self-explaining but they use quite a lot of resources for not providing great results, as the blob detection.


#3

So if you’re having trouble with, say, wind or insects triggering false positives, I take it that the “BlobDetection” algorithm is the way to go? (I don’t see the person-detection or face-detection algorithms listed in the current 1.2.0 RC.)


#4

I have also problems with wind that causes wires, leaves, water surfaces … to move and if I understand correct frame detection is the sum of changes that makes the motion system to trig. I will certainly try to use blob detection since I have several megapixel cameras that generates a lot of GB with unnecessary files. Motion View is an excellent toll to see and understand what is going on! Actually it was quite fun ;).

Blob detection was not popular in the program. It created an error - Error in Image Processing for Motion Detection Exception Detail:The type initializer for ‘OpenCvSharp.CPlusPlus.NativeMethods’ threw an exception.


#5

Hi @smithkl42 !
False positives during motion detection are usually because of background jpeg noise. Try playing with stream quality, color/exposure/white balance settings , jpeg quality settings - in your camera settings. When you have low noise picture, adjust threshold settings in motion detection so it doesn’t detect small objects like insects ore leaves. Also, if you have areas with constant movement which shouldn’t be detected - exclude it in MD settings.
Also, you may have to change how camera is installed. E.g., I had a problem with one of my cameras, I even had to exclude large area from MD because it sees window and at night car lights are triggering motion sensor every minute.
Thanks!


#6

It’s the leaves and shadows that are playing hell with my motion detection during the day, and spiders and their webs that are playing hell with it at night. For example:

I’ve keep playing with the settings, but for now, just know (as I’m sure you do) that improvement along these lines would be great. It seems to me that more intelligent motion detection is probably the next biggest key differentiator of these systems. (It’s certainly what appealed the most to me about the Nest system - but they don’t support outdoor cameras.)

Ken