How do I send less pictures to Platerecognizer

If the goal of the setup is to send to Platerecognizer only the license plates that move in a selected area in the selected direction, right?
In this case the scheme that you have now might be improved in the following ways:

1. If you have the Cross-Line Detector module is after the ANPR – this is great if you need to save only episodes of where LPs are moving in the right direction. BUT all LPs are sent to Platerecognizer in this setup, and filtering by direction happens after we get the response.
If it’s critical to filter our direction before sending LPs to Platerecognizer then we’d need to redo the scheme placing the Cross-Line Detector before ANPR. However, the Cross-Line doesn’t work without a module that tells it about the object type (in the scheme where Cross-Line is placed after the ANPR this role is played by the ANPR) so we’d need an Object Recognizer module where you need to tick object type (car). OR instead of the ‘Object Recognizer’ module you could use ‘Object Detector’ module – it reacts to any objects, no matter their type – but we don’t really need it in here.
In Cross-Line place the line with the direction going in which you want to “catch” cars.

2. If image crop is needed, you can use the Image Crop module, it is better placed after the camera in the chain.

3. Important ANPR should have these settings:
Recognize only changed areas: ticked
Detection Interval: higher than the default 50 ms (=0.05 of a second, default timeout). Try different values to find the best – to not miss anything (lower than 2000ms) and yet to not send too many pics to PR (higher than 50ms).

Please feel free to contact us if you need assistance regarding Xeoma!