Read vehicle license plates with Plate Recognizer ALPR which offers free processing of 2500 images per month. You will need to create an account and get your API token.
This integration adds an image processing entity where the state of the entity is the number of license plates found in a processed image. Information about the vehicle which has the license plate is provided in the entity attributes, and includes the license plate number, region/country, vehicle type, and confidence (in a scale 0 to 1) in this prediction. For each vehicle an platerecognizer.vehicle_detected
event is fired, containing the same information just listed. Additionally, statistics about your account usage are given in the Statistics
attribute, including the number of calls_remaining
out of your 2500 monthly available.
If you have a paid plan that includes MMC (Make/Model/Colour) data you can received the orientation of the vehicle in the entity attributes.
You can also forward the LPR results straight to ParkPow, a parking management software and sister-company to Plate Recognizer.
If you have a local SDK licence, you can optionally specify the server address.
Note this integration does NOT automatically process images, it is necessary to call the image_processing.scan
service to trigger processing.
Place the custom_components
folder in your configuration directory (or add its contents to an existing custom_components
folder). Then configure as below:
image_processing:
- platform: platerecognizer
api_token: your_token
regions:
- gb
- ie
watched_plates:
- kbw46ba
- kfab726
save_file_folder: /config/images/platerecognizer/
save_timestamped_file: True
always_save_latest_file: True
mmc: True
detection_rule: strict
region: strict
server: http://yoururl:8080/v1/plate-reader/
source:
- entity_id: camera.yours
Then, restart your Home Assistant
Configuration variables:
-
api_key: Your api key.
-
regions: (Optional) A list of regions/countries to filter by. Note this may return fewer, but more specific predictions.
-
watched_plates: (Optional) A list of number plates to watch for, which will identify a plate even if a couple of digits are incorrect in the prediction (fuzzy matching). If configured this adds an attribute to the entity with a boolean for each watched plate to indicate if it is detected.
-
save_file_folder: (Optional) The folder to save processed images to. Note that folder path should be added to whitelist_external_dirs
-
save_timestamped_file: (Optional, default
False
, requiressave_file_folder
to be configured) Save the processed image with the time of detection in the filename. -
always_save_latest_file: (Optional, default
False
, requiressave_file_folder
to be configured) Always save the last processed image, no matter there were detections or not. -
mmc: (Optional, default
False
, requires a paid plan with the MMC (Make, Model, Colour) feature enabled.) If enabled returns the orientation of the vehicle as a separate attribute containing Front/Rear/Unknown. -
detection_rule: (Optional) If set to
strict
, the license plates that are detected outside a vehicle will be discarded. -
region: (Optional) If set to
strict
, only accept the results that exactly match the templates of the specified region. For example, if the license plate of a region is 3 letters and 3 numbers, the value abc1234 will be discarded. For regions with vanity license plates (e.g. in us-ca), we do not recommend the use of Strict Mode. Otherwise, the engine will discard the vanity plates. -
server: (Optional, requires a paid plan Provide a local server address to use On-Premise SDK
-
source: Must be a camera.
If you have configured watched_plates
you can create a binary sensor for each watched plate, using a template sensor as below, which is an example for plate kbw46ba
:
sensor:
- platform: template
sensors:
plate_recognizer:
friendly_name: "kbw46ba"
value_template: "{{ state_attr('image_processing.platerecognizer_1', 'watched_plates').kbw46ba }}"
Depending on your license plate, you may recieve an template error due to variables not being able to start with a number. if so, here is another method to create the template sensor:
sensor:
- platform: template
sensors:
plate_recognizer:
friendly_name: "kbw46ba"
value_template: "{{ state_attr("image_processing.platerecognizer_1", "watched_plates")["kbw46ba"] }}"
Checkout this excellent video of usage from Everything Smart Home