Рет қаралды 527
For further information, please visit:
adaptiverecogn...
adaptiverecogn...
Images with reasonably good spatial resolution. On the license plate the minimum character height is 16 pixels for Latin and 20 pixels for Arabic characters.
With reasonably good sharpness, with reasonably high contrast, under reasonably good lighting conditions, in a reasonably good position and angle of view.
This image demonstrates an optimal image for ANPR purposes. The key point is to set the proper plate character size and to avoid problems that we will demonstrate on the following images.
In this image the number plate is overexposed because the sunlight is coming directly from behind the camera causing a burnout effect on the plate.
If it is possible try to place the camera in a position where the sunlight is coming from the side.
A similar problem may occur if the sunlight comes from directly behind the car causing a darkening effect on the car which causes is the plate be shattered or not to be seen.
There is another typical mistake that is when the camera sees the horizon. That means that either the camera position is not high enough or the camera is not tilted down enough.
This image was taken with the camera facing the sun. That causes this halo effect and makes ANPR impossible in most cases. Again you can fix it if the camera is positioned a bit higher and you tilt down the camera more.
The last typical position error is when the camera is facing the headlights. The same solution can be applied as well as in the former solutions by putting the camera in a higher position and tilting it down more.
Selecting Preset automatically sets the camera to the correct settings for the installation environment. Obviously, the settings can be further fine-tuned for the actual installation.
Surveillance
ANPR for slow traffic
ANPR for stop and go
Or default settings
The camera, which acts as the image source for OCR can be installed in many ways.
Multiple installation positions exist that can help in providing optimal images with minimal distortion.
However, there are many cases when the camera cannot be installed in these positions, causing the license plate to become distorted on the image.
This is why the OCR engine was built to tolerate a 30 degrees angle.
However, if the pan or tilt angles of the camera exceed 30 degrees, distortion may significantly reduce recognition and performance rates.
The plate on these images, for example, is too distorted, therefore Carmen accuracy might be lower than usual.
These plates however are only minimally distorted, so Carmen would most likely easily recognize them.
Lateral position
The height is minimum 1.5 meter (1 meter + the level of the vehicle headlights).
The recommended distance between the camera and the license plate is 1.5 meter to 15 meters.
Distance from the side of the road: as close to the road as possible.
The vertical rotation maximum angle ≤ 30 degrees.
The horizontal rotation degree should be lower than 30 degrees.
Transversal position
The hight is minimum 2.5 meter, with no objects blocking the camera's view of the vehicle.
Consider IR light range when determining the distance between camera and the vehicle.
The vertical rotation maximum angle ≤ 30 degrees.
The lower the angle, the better the recognition (the lower the distortion).
The horizontal rotation degree should be lower than 30 degrees.
Overhead position (in the middle of the lane)
The vertical rotation maximum angle ≤ 30 degrees.
The height is minimum 2.5 meter.
The recommended distance between the camera and the license plate is 1.5 meter to 15 meters. The Einar telezoom optics version is more than 20 meters.
To learn more about EINAR, check out our entire video series:
• HOW-TO VIDEO SERIES - ...