Human motion detection

michealkc

Hello.

I have picked a college project that proposes to be placed on a farm bull. This device would (1) Detect a human if in close proximity.(2)Generate a signal when this situation occurs and (3) Send the signal to actuate some type of moving device that would cover the animals eyes and therefore not allow the animal a chance to attack/kill the person. I have been looking online for robust human detection cameras but they dont seem to be available. Is there anybody out there that might have some advice?

ressem

The cheap option is a PIR, as used in common motion sensors, but that can't tell the difference between a person and a cow (or pile of fresh manure).
http://www.adafruit.com/products/189

If you only need a proof of concept, the closest to an off-the-shelf shape detection one that I know of would be the microsoft kinect.
Would take a fair bit of work to make the detection reliable while strapped to a moving bull's head.

A Kinect 360 teardown gives an idea of the stuff that is involved.
https://www.ifixit.com/Teardown/Xbox+360+Kinect+Teardown/4066

There's a fair amount of complicated software involved in getting this working.

Simplified a lot by using the Kinect or LEAP SDK
https://dev.windows.com/en-us/kinect/develop

chrismon

Like ressem said, pir would be the easiest option but it couldn't differentiate between cows and people.
You could wire it up to a Raspberry Pi or an Arduino and write up a bit of code.
Are you building it to actually attach to a bull or just prove it works?

ressem

Mouser have
http://www.mouser.ie/ProductDetail/Omron-Electronics/B5T-E-001-SG/?qs=sGAEpiMZZMsnnVx9hWj0EUfzsrfkWpo1ql%252bf1mI%2f9no%3d

A 6 cm x 4 cm development board costs €700. So not really student project material.
You'd need software and electronic expertise to make anything of it.

https://www.omron.com/ecb/products/mobile/okao01.html

CreepingDeath

michealkc wrote: »

Hello.

I have picked a college project that proposes to be placed on a farm bull. This device would (1) Detect a human if in close proximity.(2)Generate a signal when this situation occurs and (3) Send the signal to actuate some type of moving device that would cover the animals eyes and therefore not allow the animal a chance to attack/kill the person.

That's a serious project with a lot of technical challenges.
You'd have to strap the electronics and battery pack to an animal which may roll over on it's back to scratch itself.
Or if you had some "GoPro" style camera mounted on its head, it would more than likely find the nearest tree to knock it off.
I'd worry about how you could cover an animals eyes, without having some form of "cow goggles" on it with some LCD display that blanks out the lense on command.

If you try detecting humans optically, then that's a complicated computer vision problem. Not including the fact that bulls have a much wider field of view with their eyes on the side of their heads to detect predators.
And I imagine if they charge at you, the head is faced downwards so the camera may not show the human in front.

If you try detecting humans via infrared sensors, you wouldn't be able to distinguish between other cows and humans, nor if the cow turns around and part of itself came into the sensors field of view.

I'd probably consider something a little less complicated.
Maybe sticking bluetooth tags/tiles on the cows, and being able to detect their presense in the field with a Raspberry Pi.
And send a signal if they start disappearing - anti-theft/cattle rustling.

I know there's dog collars which give an electric shock/or uncomfortable noise if they stray outside their garden. That might be used to modify the bulls behaviour, that it goes off if a person comes near them with a special transmitter. If effect you would have a virtual electric fence.

If you could do that right, then there might be no need for fences around animals at all.

But if you need to simplify the problem of detecting a human, I'd start by assuming that the human has some form of transmitting device on them eg. bluetooth/wifi/radio etc. And the bull has some harness around their neck, that emits some sound/mild electric shock.
Although that would likely condition the buill to fear humans, which may not be want you want, eg. if a vet is trying to examine them.

That's my 2 cents for this interesting challenge.

CreepingDeath

Another possibility, is to have a 3-axis accelerometer sensor on the bull, and detect when it is moving fast and assume that is an attack/it is running.

That could have a lot of possibilities, including detecting when an animal isn't moving as much as normal over the course of the day/week (may be ill).


FYI: I've been waiting for the Wimoto bluetooth sensors to release. (waiting over 6 months already for general release, but haven't ordered/pre-ordered any)
Bluetooth devices running on a coin battery for up to a year.
The "sentry" one has an IR sensor an accelerometer, but hasn't been released yet.

http://www.amazon.com/Wimoto-SEN001-Sentry-Bluetooth-Sensor/dp/B00ES5BJOO/ref=pd_sxp_redirect/188-9654571-7167122

michealkc

Thanks for your reply. This device needs to work in concept. Then if all that is ok we try to make the thing. If it is successfully made we try it out for real.

michealkc

Thank you for all that information and I will read carefully over it as I am very keen to make some type of breakthrough. You have certainly given me some valuable advice. You sound quite knowledgeable.
Unfortunately passing electric shock onto an animal seems to be against the law. Which changes the scope of the project.

CreepingDeath

One thing that could be a starting point for you is face detection on a Raspberry Pi. The benefit of Raspberry Pi is that you can power it with a portable "power bank"/battery pack, and it already has IO ports to connect a light/buzzer/electric relay.

Here's an rPi project that detected faces.
But for your project that's not the same as human detection, and wouldn't detect the bull creeping up behind someone with their back turned

https://www.cl.cam.ac.uk/projects/raspberrypi/tutorials/robot/face_detection/

But not a bad starting point.
Ideally you want the later Raspberry Pi 2 version, with the quad-core processor.