Just wanted to ask the community for ideas on detecting, with decent certainty, that there is a car standing in one of the two spots in may two-car garage.
Have been thinking of sensors based on lights or ultrasound - but that can be triggered by someone just walking by.
One idea is the sensors that is used for traffic lights in some countries in Europe (detects presence of a large magnetic object?).
What are your thoughts?
Do you need to detect their position, or just their presence in the garage? If it’s only presence, you can put a Bluetooth beacon in each car (or an old Mi Band, since you can connect them to USB to avoid changing batteries). Then use something like and ESP32 or a RaspberryPi to detect the beacons.
Maybe place a bluetooth iBeacon in the car and use an ESP32 board to detect the beacon?
I use one to detect that my bike is in the garage. The iBeacon reports the distance from the receiver with quite good accuracy.
https://www.amazon.co.uk/gp/product/B07WBXK2W4/ref=ppx_yo_dt_b_asin_title_o02_s00?ie=UTF8&psc=1
This works well, I use iBeacons with a pizero w and room assistant to track the presence of vehicles.
I like the Beacons and am planning to use them in some other projects. It might work in the specific scenario for specific cars, but I wanted to challenge this a bit by beeing able to do this for any car.
A cheap camera and Deepstack object recognition?https://community.home-assistant.io/t/face-and-person-detection-with-deepstack-local-and-free/92041
I have this set up and the resulting sensor reports the number and type of objects detected. Around 40 different objects can be detected.
I suspect that this is the way to go. I suspect that they use that kind of technology in large parking garages where they indicate a free spot with a green light, and a taken with a red.
Any other ideas?
You could detect the license plate also to see which car it is.
This is the output from one of my deepstack sensors. It is looking for persons but in this case,the last seen object was a chair. It also saves a picture in a specified folder with the detected object framed.
target: person
target_confidences:
summary: {
"chair": 1
}
predictions: [{"confidence": 60.6, "label": "chair", "box": [0.642, 0.562, 0.844, 0.637], "centroid": [0.6, 0.743]}]
last_detection: 2020-01-03-18-31-35
save_file_folder: /config/www/person_kitchen/
friendly_name: person_detector_kitchen
How about an esphome using a hall sensor to detect the magnetic field disturbance of a car? You could place them under the front-middle of the car spots (or maybe hang them down and make them act like “stop lines”)
Cool - got some intial ideas, and I really liked the license plate one. Might post it in the project forums later.
Normal and “professional” products in large parking areas are for example this:
And are always using a ultrasonic sensor.
If you are consideration of build something, here two projects to start:
I think bad idea is here spent time with Chinese products. 
I created such sensor using Fibaro Smart Implant (in analog voltage mode) and sharp infrared distance sensor. Sharp sensor is powered by 5V and on output it gives voltage is is proportional to distance from obstacle (0.2~4.5V). This voltage is measured by Smart Implant and gives good indication if car is in garage. Sensor is suspended over the place where the car parks. There are different versions of Sharp IR sensor available, with different distance measuring distanc ranges, so it is easy to select one that would suit your your particualr environment. All what is needed is working Zwave network (for implant connection), sensor and power supply (I sue 12V for implant and 5V for sensor). Installed ~6 months ago and so far 100% reliable.
I went with this route for our two car garage
Some of you guys haven’t read the question. @Koffeinuser doesn’t want to use ultrasound for fear of false triggering. However it could still work if you filter out short periods of detection so eliminating people walking past.
A combination of sensors coupled with a bayseian sensor to combine them?
Or using multiple sensors looking in specific detected ranges for the signature of a car.
Your’e almost getting into the photo recognition stuff at that point!
I use an ultrasonic sensor attached to a D1. It checks every 5 seconds (user defined). The delay tends to filter out the stray transit of the garage by a humanoid.