LPR Overview
License Plate Recognition (LPR) is another commonly requested video analytics feature and serves a similar function as Face Recognition in that it detects uniquely identifiable features of a car but instead a face it uses a license plate.
License Plate Recognition is used often in real-world situations, including, but not limited to:
- Parking Management
- Traffic Violations
- Stolen Vehicle Recognition
- Logistics / Shipping Tracking
Generally, LPR systems produce several probable license plate numbers with varying levels of confidence. So while tracking a license plate on a video stream it's good to keep track of probabilities in order to return the most likely result. It is also ok to send all options recognized above a certain threshold, depending on the accuracy of the integrated solution.
Detected LPR Object Format
- Type: Car
- Attributes:
- License Plate Number
- License Plate Number alternatives (e.g. LPR1 % confidence, LPR2 % confidence)
- License Plate Country
- License Plate Region/State
- Additional vehicle data (e.g. Color, Make, Model, Year, Speed)
Recommended LPR Event
The License plate detected event should be sent once per track, so it is ok to wait for a second or so to determine the most probable license plate number. Caption: {License plate number} ({license plate number candidate 1}, {license plate number candidate 2}). Description: {all car attributes}.
Vehicle Database Integration
In some projects it is convenient to support an additional vehicle database to keep information associated with the specific car, for example, the driver's name, phone number, list name for the vehicle.
In this case, you should send all that information with detected objects and detected events.
Comments
0 comments
Article is closed for comments.