Best discreet/hidden motion sensors for outside front of house?

[NeverDie]

It seems like there's a lot of very powerful free open source video analytics software out there, like OpenCV for example. Has anyone here tried OpenCV, or something of at least equivalent strength?  At least during the daytime, is there still nothing useful in that category that would sort this?  Most of these youtube videos look real world.  Are they not?  i.e. Are they the real deal, or are these youtube videos (below) just selective edits of the few times it happens to be working as it should?  For instance, I don't see much shadowing in these videos, so perhaps that real world element fouls things up?
 
As discussed in another thread on cocoontech (http://cocoontech.com/forums/topic/26642-diy-gasket-material-for-outdoor-video-cam/?p=214718), I have a 3MP outdoor videcam on order.  It will even be much more compact and less conspicuous than the patheticly larage  Visonic Opto-Quad tower above. It  can allegedly record 3MP video at 20fps or 1080 at 30fps.  This stuff is getting cheaper and cheaper while getting better and better all the time.  Seems like there should be a way to leverage it for the comparatively modest purposes of the original post on this thread. 
 
At least with a video image, it should be a lot easier to mask off (in software) areas you want to exclude, unlike PIR's where the masking is physical and comparatively primitive.
 
 
 

https://www.youtube.com/watch?v=LNaL58GRou4

 

 

[NeverDie]

https://www.youtube.com/watch?v=InqV34BcheM
 
https://www.youtube.com/watch?v=4jKyzFE9Dy0
 
 

 

[Del91574]

Without a smart back end I don't think you want to get involved with real analytics. Falses are no different than a poorly installed detector.

 

[NeverDie]

I found this IP66 ceiling mount 360 outdoor PIR:    http://www.sensorswitch.com/DataSheets/SBG-6&10-(OEX).pdf
 
"Gasketed for use in a Wet Location Luminaire".  Seems like it might work.
 
It's merely the first one I've found.  I have no familiarity with the company, and I have no idea what reputation it has, if any.  So far I've been unable to locate a source to buy it from.
 
Maybe there are some other kinds of wall or ground mounted sensors that could blend in, depending on circumstance, but I haven't seen any yet.
 
I'm surprised there aren't more COTS options and alternatives.  I guess this means that most people, if they have any front perimeter sensing at all, are using massive sensors that don't blend in and/or are mounting them in ways that pull attention toward them?
 
Sounds like that's all she wrote.

 

[Del91574]

Link is dead.
 
Ceiling mount PIR or luminare style is going to false...think about what it's looking at and what is likely to cross the ground within it's view.

There's options out there for detection but anything outdoor is also going to be likely to have many falses and enviromental concerns. I installed ground based radar units and leaky coax and fiber detection in addition to MW detection for air base flight lines and runways.....there's ways to do it but you're not going to like the price.

 

[NeverDie]

think about what it's looking at and what is likely to cross the ground within it's view.

 

A stray animal?  Leaves blowing in the wind?  The irrigation sprinklers?  
 
Hopefully this link will work:

http://www.sensorswitch.com/DataSheets/SBG-6&10-OEX.pdf
 
I'm not disagreeing with you about the concern you raise, but I do notice it claims "The sensor has special outdoor PIR detection settings (OEX). These settings ensure that environmental factors, such as wind, do not cause false ons."

 

[Del91574]

Depends on your application.
 
Unit is designed for lighting automation and control.
 
Animal and leaves would be what I'd look at. 

 

[NeverDie]

Seems like a 3D sensor could overcome those types of false alarm scenarios because--at least in theory-- it could screen out objects too small to be humans.  This soon-to-be-released 3D sensor is priced at $350-$500:
 
http://structure.io/
 
but it's presumably also a lot more accurate than what's needed for the purpose indicated here.  Kinect has a friendlier pricepoint.  Has anyone tried adapting a kinect for front yard perimeter security or similar?
 
Not sure what sort of outdoor enclosure would be needed for it though.  

 

[NeverDie]

For those with interest, I just noticed Skanect--the same software used by structure.io above--is free for personal use and it works with Kinect....

 

[Del91574]

From dealing with real analytic systems (ala Genetech, Verint and others) there's a limitation in functionality in all analytic systems to detect a person specifically and trying to do such, when you start considering distance and the resolution of the native system, they don't work.
 
The analytic portion is used to set up a virtual tripwire or to generate a baseline of what should exist in the background and what should not (obscuration or removal of something). I'm using 3-5 mp cameras on the front side, compared to a relatively simple kinect sensor and close FOV.