I'm planning on making a "Sentry Mode" like system as Teslas have so when the car is parked and something moves in front of a camera, the clip is saved into an SD card.
I plan on having 4 cameras. My first question is how much will it draw at "idle" which means detect movement but not write on the SD, and how much will it draw while writing to the SD.
Next question is, my power supply will be 12 volts so that I can have longer cables as the cameras will be positioned far from each, which induces the next question.
Can Arduino power 12v camera modules? (If such modules exist ofc)
Another solution is to have 5 arduinos. 1 main and 4 far from each other next to the cameras so I can use 12v to power the Arduinos that are far from the main while bringing power to the cameras. But this would be more costly and probably overload the main board...
Any guidance or change is welcome. (Might throw some wifi connectivity in there later).
The power draw will depend on many things, among which is the type of Arduino you plan to use. If you did not realise that all Arduino are not the same, you have much research to do, to find an Arduino capable of detecting movement and saving a "clip" (I assume this means a video clip, not a still image) to an SD card from a single camera, let alone from 4 cameras.
The initial thought is 4 of these modules: ESP32-CAM-MB 5V connected to an Arduino uno.
But if there's better alternatives I'm here for it.
I calculated roughly that this will draw at the worse case 0.036kwh if ran for 24 hours at max for one camera, so 4 cameras running at max would be 0.144kwh for 24 hours. I'm mainly asking for guidance on how to best do it. Especially with the cables cuz 12v is a must in such distances.
I got people detection working with a IR camera. I have a ESP32 running an IR camera that transmits its image to a Raspberry Pi that runs a program using Machine Learning/Tensors to detect people and small animals. An Uno is nowhere near up to the task of image processing and object identification. An ESP32 can run simple ML and sort through some small Tensors. A ESP32 can do KNN (Nearest Known Neighbor) ML and Linear Regression ML.
I'm more interested in general movement, not specific object identification. I wanna record anything that moves in front of the camera, be it human, cat or some flying object, Which should be way easier and less demanding than running ML. I'm fairly confident an uno can do image movement. Haven't tried it obviously but I can't see why not. Do you think generally a raspberry pie would be better for this? I wanna keep it low cost as much as possible.
Imagine if you will a 8 bit machine with 20K of ram doing image processing on a 640x480 color image. Even if the image was cut down by 20% the Uno will have how much RAM remaining to process the image?
I think an ESP32 CAM, and another ESP32 for image processing could do the trick. I am doing weather predictions with a ESP32, it works pretty good giving results for the next 24 hours.
I'm trying to do the exact same thing but i have 1 ESP32 cam with a fisheye lens and I'm doing image prossessing onboard. Check out this project using an ESP32 as a dashcam which was really helpful. I'm just going to roll the footage (delete least recent) if it ends up filling the storage and dump it to a pi i have on my home network every time i get in range. I park on the street right outside my house so this should hopefully have decent connection (yet to test this...). In terms of power, be careful taking directly from the battery, take it from a connection that cuts out when the engine starts else the power spike will kill your 12v to 5v converter the first time you start your car. I suggest the cig lighter or radio power. Good luck and keep us updated with how it goes!