Suggestions for a DIY perimeter alarm system

[tsi]

Hi all,
 
I am new to the forum, coming from the SmartThings community. I am into home automation, but 
I have been thinking about a perimeter alarm system lately.
 
Smartthings is great for home automation (and playing around) but i regret to say that the cloud dependency and general instability are major showstoppers for security purposes. 
 
The project i am investigating includes the following key parts:

1. A number of outdoor IP cameras (ideally hardwired via Ethernet). IP E27 wifi bulb cameras have crossed my mind (minimal installation -  low profile) but i haven't found any supporting ONVIF/RTSP.
2. A number of z-wave motion sensors (are there any outdoor ones?) Or are the ones on the cameras themselves sufficient?
3. A home automation system that RUNS LOCALLY, such as indigo, OpenHAB, home-assistant (suggestions are welcome).
4. A z-wave USB stick (suggestions are again welcome)
5. Sighthound software running on a capable Mac/PC which is connected to a UPS.
6. One indoor (Aeotec) and one outdoor (Popp solar powered) siren. Both are z-wave.
7. Two z-wave open/close sensors.

What i want to create is a simple but efficient perimeter security system for my home.
 
I will not get into trivialities like the number of possible entry points and such. For simplicity's sake let's just assume that i only have one entry point that i need to secure.
A single IP cam is looking over the area of interest and Sighthound can detect the presence of a human form. The motion sensor as well can register the thermal signature of a possible intruder.
 
A very simple rule runs, as follows:
IF Sighthound detects a human form AND the motion sensor registers something THEN sound both sirens at max volume.
IF Sighthound detects a human form OR the motion sensor registers something THEN sound the internal siren at low volume.
In both cases, Sighthound sends rich (with photo) notification of the event via e-mail and ideally via push notification (iOS) as well.
 
What do you people think about the above setup? Would it work? Would it be reliable?
Sighthound can run external commands in case of an event, by the way.
I am eagerly awaiting your comments/observations/objections.
Thank you very much in advance.

 

[BraveSirRobbin]

Interesting setup.
 
I would look into Dakota Alert's perimeter motion detection, especially if this is going to be a large detection area.
 
http://www.dakotaalert.com/catb2b1/product_info.php?cPath=51&products_id=143&osCsid=kgrng5m0oot00thsu4gqakish3
 
You can monitor contact closure for further action with your system.

 

[cobra]

Outdoor motion sensors tend to be very false alarm prone due to the potential for animals and wind driven triggers.  Most lower end sensors will not be usable outdoors.  There may be some more expensive outdoor units that can discriminate better based on human profile, but I suspect the camera based motion detection may work better (although this may be error prone as well.)

 

[Del91574]

Outdoor motion sensors tend to be very false alarm prone due to the potential for animals and wind driven triggers.  Most lower end sensors will not be usable outdoors.  There may be some more expensive outdoor units that can discriminate better based on human profile, but I suspect the camera based motion detection may work better (although this may be error prone as well.)

Camera based tends to have a lot of other issues, even when strong analytics are used, which rules out the lower end systems, even if custom scripts are used.
 
Take a look at this:
 
https://www.youtube.com/watch?v=T2cEIUTiq5w
 
Pay close attention from about 1:30 on. The analytics shift from the primary target to a different (possibly incorrect) target due to the contrast and amount of pixel change of the primary target vs. background. No fault of the analytics, but the contrast between the foreground object vs. background object and analytics attempting to make the best decision.
 
Outdoor detection is an art upon itself which requires a lot of planning and consideration, not something that is normally thrown together with basic items and hoping for the best.

 

[tsi]

Hi all and thanks for getting back. My house is on the second and third floor, therefore my perimeter comprises of some balconies. It is (I think) a rather easy setup. If someone shows up after I activate the alarm, then it is an intruder wannabe. Simple as that. What I want to rule out is falsely triggering the perimeter alarm because of an animal or plant or clouds or smth.

I am fairly confident that the computer vision algorithm will work most of the time but I am afraid of false positives. This is why I am thinking about the motion sensors as well as the cameras.


Sent from my iPhone using Tapatalk

 

[dementeddigital]

Both outdoor motion sensors and camera systems have the potential for false positives.  Using both in conjunction with each other seems like a good idea against false positives, but it also creates another failure point since both devices have to trigger. 
 
I had some PIRs outside years ago, and every so often (usually at 4AM) it would trigger like crazy because some insect was doing endless circles around the sensor.  Depending where you live this might not be a problem, but here in FL the insects are plentiful.  You'll want to get a PIR made for outdoor use.  Maybe you can find one which is a combo PIR/microwave device.  You'll want to mount it somewhere where it won't get any direct sunlight.  Test it so that it is sensitive and has no blind spots, since your alarm requires both this and the camera to trigger.
 
I hadn't seen Sighthound before, but it looks pretty slick.  I have lots of trees, and they cause video triggers on my simple Q-See system when I don't want them to trigger sometimes.  Sighthound seems to have a trial version, so I'd recommend downloading that and giving it a try to see how good the human detection works where you need it to work.  Put the camera where you are going to finally mount it for the testing, since the camera angle might create problems with their person detection algorithm.  Try it at various times of day (various light levels).  Can it mask out the detection of a person walking outside on the first floor?  If you don't mind posting back, I'd be very interested to hear how well it works!
 
Have you thought about any other kinds of sensors?  A simple lightbeam might work well.

 

[tsi]

Hi again,

Regarding Sighthound I have done some preliminary testing and the human detection works.

You can define certain areas that the software is processing for motion.

In my use case that should be enough because my house is in 2nd and 3rd floor therefore anyone on the balcony (after i arm the system) is an intruder who climbed the wall or the adjacent building.

There are some advanced features like "detect a person who had entered a door or crossed a line" with both door and line configurable by the user, however my test were mixed.

I live in a rather suburban environment and I am mostly interested in the night time (infra red) performance of Sighthound.

Regarding the motion detector redundancy mechanism I am still confused.

I am still in the early stages of evaluating whether the setup I propose is a good idea or not.

Cost is also a consideration. Such a setup (6 cameras) requires a dedicated server with quite some punch. i7 with 16 Gigs of RAM might do the magic. A UPS is also required.

I welcome any suggestion, comment, objection regarding the project.

Thank you all who have participated.


Sent from my iPhone using Tapatalk

 

[RAL]

Since they are second or third floor balconies, I think I would simply put a contact on the doors/windows and be done with it. Skip the cameras and motion detectors.  You're spending a lot to guard against the unlikely case that someone actually is able to get up there.