Recognition Settings

Recognition Settings

To obtain ANPR results you will need to:

  • connect to the video sources

  • adjust recognition zones

  • adjust ANPR analytics

It is also important to have the connected cameras mounted and set up properly.


ANPR Setup

Connect video source

Navigate to Settings > Connection. The tab contains 3 blocks:

  • [Choice of video source]

The number of connectable video sources (channels) available is determined by the license.

  • [Video source] connection

  • Recognition system block

Click APPLY button for the changes to take effect.


Choice of the video source

In a vertical panel on the left please choose an appropriate video source.


Video source connection

NumberOk supports three types of connectivity:

  1.  RTSP connection.

Sample  of RTSP string to connect to a AXIS IP camera:

rtsp://root:pass1234@10.0.3.170/axis-media/media.amp, where:

root – is a login of the camera user;

pass1234 – is a password of the camera user;

10.0.3.170 – is a camera network IP address;

Sample  of RTSP string to connect to a Hanwha IP camera:

rtsp://admin:PassWord123@10.0.3.130/profile2/media.smp, where:

admin – is a login of the camera user;

PassWord123 – is a password of the camera user;

10.0.3.130 – is a camera network IP address;


  1. CAMERA/DVR connection is based on video sources connection through SDK provided by respective vendors. NumberOk supports the following connectivity methods:

o      DVR Hikvision;

o      DVR Dahua;

o      DVR NOVUS B series;

o      DVR LINIA;

o      DVR TVT;

o      DVR Praxis;

o      DVR NUUO;

o      DVR Partizan.

Input fields:

o      Login – login of DVR user;

o      Password – password of DVR user;

o      Address – IP address of the DVR;

o      Network port – network port of the DVR;

o      Channel – DVR channel (camera) number;

o      Stream – DVR stream (main/sub, 1/2/3 …) DVR.


  1. Video files.

Video files are useful for setting up and testing NumberOk. The connected file is played in a loop. Also, a video stream URL can be specified as a file source (i.e. video stream from the camera). This method applies to the video stream of some special normalization tools and can be used as an alternative way to obtain a streaming video.


Choose the type of connection, input its parameters, and push the Connect button. Provided the source is available and its settings are correct NumberOk will confirm the established connection.


Main video sources supported by NumberOK:

RTSP stream  with H264 codec

HTTP stream with MJPeg codec

image.png

Local video files:

        - Codec: H264;

        - Container: AVI, MP4, TS.


Video stream parameters:


Resolution  - from 1 MP to 8 MP

Bit rate  - any restriction

Frame rate  - 1-60 frames/s


NumberOk already supports the H265 codec. If you want to connect cameras with the H265 codec please insert the RTSP string into the "File" field.

We recommend using libraries for decoding  ffmpeg4 and live555

 

Connection Troubleshooting

If NumberOk reports video source/stream unavailable, please, check the following:

  • video source URL is correct (protocol, IP address, port, parameters);

  • video source (camera) is actually available (e.g. try connecting using third party software, e.g. VLC);

  • camera username and password do not contain special characters (e.g. / : ' " \ @ ) as these have special meaning when used in URL or are URL-encoded;

Character

Replace with


Character

Replace with

@

%40


#

%23

:

%3A


'

%27

%

%25


"

%22

/

%2F


\

%5C

Please, use any public service (e.g. https://www.urlencoder.org/) to ensure your username and password do not contain any special characters;


ANPR Settings

Recognition Zone 

Recognition Zone is a particular zone within the frame the number plates are expected to appear in. No recognition zones are activated by default.

Set the zone boundaries by dragging the vertices of the visualization quadrangle. The shape of the zone should be so that the number plate remains within the zone for as long as possible. Keep the zone's sides parallel to the frame’s boundaries whenever possible. Avoid OSD and regular structures (e.g. fences). 

You can set up to 4 zones within a single frame to make a distinction between e.g. lanes, entry, and exit stop spots etc.

It is worth placing the zones where the number plates are well-lit and read. It is strongly recommended to avoid zones overlapping.


Number plate size

NumberOk will search for and recognize number plates with widths within the specified range. The minimum permissible size of the number plate for recognition is 130 pixels, and the maximum is 1000 pixels. By default, the range is pre-set at 130..250 pixels.

Number plate size in the frame in full resolution can be measured by means of common image editors. Alternatively, you can pre-set an obviously wide range and analyze the dimensions of the recognized number plates in technical info on the Results > Recognition Results tab, you can also set up a narrower range based on technical info to improve overall system performance.  

Please, set the range as close to real values with a small add-on. Too wide a range increases the load on the processor and affects recognition timing and quality.


Results > Recognition Results mode reflects the width of the recognized number plate in pixels (px).


Recommended range from this example is 160..270 px.

Recognition sensitivity 

Each number plate is recognized with its own ‘Confidence’ level. The better the number plate image quality the higher the ’Confidence’ level. Accordingly, false number plates in the frame (the recognition results that are not number plates) will have a lower ‘confidence’ level.  To cut off the ‘garbage’ from true number plates you choose the threshold of confidence higher than garbage readings and lower than real ones. The threshold of confidence (50% by default) is strongly dependent on video stream quality. For more confident ANPR you should raise the threshold up to 65-70%. In the column of the results technical info in the Results tab you can find out the real values of recognized number plates ‘Confidence’.

Values below 0.65 (65%) for real plates readings indicate that one or a number of the following should be improved:

  • image quality (exposure);

  • lens focusing;

  • mounting point distance and view angles;

  • video stream quality (bitrate, fps, i-frame, resolution) 

Results > Recognition Results mode exposes the confidence for each recognized number plate in 0.хх format.


This example contains values below 0.65, which means that video image quality improvement is required. If the confidence level is set at 70% (0.7) number plates with confidence values below 70% (0.7) will not be included in the recognition results.

Direction angle

Specify indicative direction to enable direction detection (in/out). The parameter value range is 0 to 359 degrees, with a step of 1 degree. Adjust the arrow by moving the slider or mouse scroll.

The arrow direction indicates the direction to enter (e.g. controlled area).

When the plate movement direction coincides with the arrow direction, the ‘Entry’ event is detected; in case of movement against the arrow direction ‘Exit’ event is detected. Any movement with azimuth diverging for more than 15 degrees from the preset direction results in an ‘Undetermined’ value.

An approximate reference angle of movement or alignment along road marking can be not sufficiently precise, notwithstanding the tolerable deviation of about 15°. It is much more reliable to pre-set reference angle of movement according to the analysis of already recognized number plate technical information.

Results > Recognition Results mode reflects the recognized number plate movement direction in azimuth degrees.


From this example the recommended direction arrow angle is at  90° for ‘Entry’ registration and at 270° for ‘Exit’ registration.


The absence of degree values indicates low frame rate, video quality, or other image or video stream quality issues.

Additional materials

NumberOk provides advanced settings of recognition to obtain the best possible results. Please consider familiarizing yourself to the LPR Analytics Setup guide.




    • Related Articles

    • LPR Analytics Setup

      NumberOk offers advanced recognition settings to compensate camera installation disadvantages, bad weather conditions, lack of lighting, and number plates defects. This manual describes ANPR Analytics Setup in details to allow the user to set the ...
    • TCP Integration

      Preparations Set up NumberOk to recognize license plates. You will notice LP overlay accompanying vehicles under View tab when recognition is tuned up properly. TCP Integration settinga Basic settings Go to Settings > Analytic. Check Analytics: ...
    • Running NumberOk as an operating system service

      There is a possibility to run NumberOk as a service to better integrate it into distributed computation systems running on server platforms. As a service, NumberOk is better integrated to a multi-threaded environment, optimized to work with TCP/IP ...
    • LPR Cameras Connection (NumberOk META)

      1. Adding LPR camera in NumberOk Meta Navigate to Settings > Integration > LPR Cameras Click Add Camera button. Add LPR camera dialog will open. Complete the dialog with the following: Camera type - FF group protocol Describe your camera Specify ...
    • Access control: Checkpoint

      Checkpoint: General Information Switch on Checkpoint mode at the tab Settings > General. Checkpoint subtab is available only in Checkpoint and Parking modes. More details on Parking mode setup: The tab is designated to configure checkpoints and ...