Lora Project help and suggestions etc.

Hi Ron and @noweare

I am sorry if I caused confusion.

Ron you are very right about the government helping on something like this. It would never happen.

Also thank you for commenting on my previous reply, I think I will tackle that project first before revoking the security issue.

So a gateway between lora and the internet could work for the tank level monitor and manual switching of the valve. and microcontrollers and Lora doing the auto.

In the meantime a few of the culprits were arrested after being caught on camera from the neighbor farm, we could identify the dogs and went looking for the dogs and consequently caught the owners.

@noweare, I am and have been flying drones for many years and drones can definitely work, especially in conjunction with another sort of security solution as my first post implies. The problem on my side is the cost at this stage, with our N$ being what it is. DJI has quite a few options for thermal imaging drones, the cheapest that I could use fairly efficiently day and night cost around 12,000USD and a little out of my budget for the time being.

I know about 1 farmer in our community, that just bought himself a DJI Matrice M30T < https://enterprise.dji.com/matrice-30> for around 12k USD

Regards all

@smiddie Check the Swiss Guy for everything LORA, and have a look at MeshTastic.

The Swiss Guy

I'd look into https://meshtastic.org/docs/hardware/devices/ for this use case personally. Probably a RAK nRF52 module

Then maybe hook up one of those Xiao ESP32-S3 cams for on device inference, there's a pre-trained model for person detection you can load from a webui.

Then you just need to link the two microcontrollers/API's & add a solar system, https://www.tindie.com/products/burgduino/mppt-psu-5-24vin-5v-3v3-output-15w-combined/ would probably be more than adequate.

Should be able to keep BOM to ~100 or so. Will probably still get stolen though..

@smiddie Very challenging problem. I wish you luck on your project and getting the bad guys. Wish I could help more.

@sam_uk @smiddie I really doubt one of those tiny 2MP cameras mounted 20ft in the air can do facial recognition from 20 feet in the air and 100 yards out, but it's cheap enough to try out. There are much better sensors and lenses available, all at a cost of course. Cost could easily be over $100 per camera, I just bought the Pi HQ and it is close to $100 already.

I want to get one of those modules to compare it to all my other options, but I note that the spec sheet says WiFi consumption is 110mA, and the camera is from 154 (NO video) to 304mA in picture mode (not sure if that is just a snap, or includes video). Simple math tells us it is consuming over 2,000mW (5V x (110ma + 304ma)) per hour. The good news is the solar is producing about 10 times that with just two 10W panels. That means while the sun is out the batteries stay at 100%. 

Doing a solar budget gives us: solar output is 20,000mW per hour. Camera is using 2,000mW per hour, so we have 18,000mW per hour (when the sun is shining***) for recharging if operating overnight (part of the night only). The empty battery pack can be 100% full after only about 7 mins.

I think my math is right, but if somebody sees an error please let me know.

You have two choices, one is to only run the camera during daylight hours and the other is to only enable the camera and WiFi periodically. There are several people here me included who know how to totally turn off the esp32 board so draw is zero. I would recommend a combination of those two strategies.

ADDITIONAL IDEAS:

I would also look into infrared cameras otherwise known as night vision. I am sure I saw them mentioned in the last few days, and of course using an infrared illuminator makes it even better.

Getting rid of WiFi and using LORA instead saves a lot of power. With directional and long range antennas, LORA can go easily over 10Kms.

By turning the camera off for 3 mins then on for 1 or off for 90 seconds and on for 30 secs (my choice) the batteries will last 4 times as long.

Good luck!

Thank you everyone @zander, @sam_uk for your inputs, this looks more and more viable to me 😏 I still need to get some hardware to start at least somewhere.

Just one thing Ron, the cameras are not to be used for facial recognition, but rather "object recognition" not sure if there will be a difference though? and distance will not be further than about 20'

I will keep updating, and please continue with ideas if you wish.

Regards

@noweare No worries and thank you for your interest...

@ron the ESP32-S3 Xiao cams can be upgraded to a 5Mpx camera module. It's probably a longshot, but it only costs $15 to find out if the person (not face) recognition is up to the task.

@sam_uk The 5MP will help for sure.

BTW, I am @zander