To ensure the recognition of state license plates using the License plate recognition (VT), it is necessary to install and configure video cameras in such a way that the requirements described below are met.
The video camera must be permanently installed.
The frame size is at least 704*288 px.
The character height for IP cameras is at least 20-30 px, for analog cameras—at least 14-20 px, the stroke width is at least 2 px. If it is necessary to recognize two-line license plates, then these requirements are relevant for each line.
The minimum allowable contrast in case of even staining of the license plate must be at least 20% (contrast visibility of characters relative to the background is 25 units with a 256-point scale).
The maximum allowable uneven staining is not more than 12% (the ratio of the stained area to the total area of the license plate).
The geometric proportions of the image must not differ by more than 10% from the geometric proportions of the state license plate itself.
For MPEG-4 and H.264 formats, the recommended distance between key frames is no more than 4-8 frames in case of a stable connection. |
The maximum vehicle speed for correct license plate recognition is limited to 180 km/h. |
The focal length settings of the camera must be correctly set. The table below provides an example of the correlation between the viewing angle, sensor size, and lens focal length. For calculations, a standard video surveillance calculator was used.
|
The camera must have a lens with an aperture of f/1.0, f/1.2, f/1.3, f/1.4, or f/1.8.
Color space model: Y800, RGB32, RGB24, YUY2, or UYUY.
The camera must have infrared illumination (built-in or stand-alone IR projector) to ensure round-the-clock recognition in an area with changing lighting conditions.
The shutter speed settings of the camera must be correctly set depending on the level of the camera's tilt to the license plate, as well as on the vehicle speed. The table below provides approximate shutter speeds for different vehicle speeds.
|
The camera must have a photo sensor suitable for license plate recognition:
Sensor type | Both CCD and CMOS sensors are suitable for license plate recognition. CMOS is preferred because it does not have the blur effect that is common with CCD sensors |
Sensor size | 1/3 inch or more. During recognition, the physical size (area) of an individual pixel in a sensor is more important than the physical size (area) of the sensor itself. Larger pixel size provides better recognition in low light conditions |
Pixel size | 3.75 micrometers (µm) or more. The larger the sensor pixel size (area), the higher its sensitivity |
Sensitivity | Dynamic range of 60 decibels or more (approximate contrast ratio of 1:1000). It is recommended to use the wide dynamic range (WDR) function. It improves the quality of recognition in high-contrast lighting conditions—it is possible to capture image details in both dark and bright areas of the field of view |
Color depth | 10 bits per pixel (bpp) or more |
When choosing a video camera, we recommend you to rely on a comparison of the photo sensor characteristics obtained as a result of measurements in accordance with the EMVA 1288 standard. This standard was developed by the EMVA (European Machine Vision Association) and it is offered to video camera manufacturers as a way to unify product evaluation methods. |
The camera tilt angles do not exceed the maximum values:
Angle | Recommended value | Maximum value | Scheme |
---|---|---|---|
Vertical angle (camera tilted down to license plate) | 18-20° | 30° | |
Horizontal angle (camera tilted left/right towards license plate) | 5-10° | 20° | |
Tilt of the license plate in relation to the road surface | 0 | 5° | During the recognition of a single-line license plate (with a minimum of six characters), the camera angle is checked by the "one-line rule": an imaginary horizontal line must cross both the first and last characters on the license plate. |
For correct license plate recognition, make sure that the image of the license plate is not:
Any deviation from the above requirements reduces the probability of correct license plate recognition. |
An example of license plate images that will be recognized correctly and completely: