Page History

...

1. Go to the settings of the LPR IntelliVision module object created under the LPR channel object.
   

2. Set the Use GPU (1) checkbox if it is necessary to use the NVIDIA GPU resources to increase the license plate recognition performance. By default, only the CPU resources are used.

   Note
   title Attention!
   The startup (initialization) of the license plate recognition algorithm on NVIDIA GPU can take about one minute. No LP recognition events will be received until the initialization is complete.

   
   

3. Select the LPR accuracy in the Accuracy (2) drop-down list. The following options are available:
   
   1. Maximum - enables maximum recognition accuracy, but it causes great CPU and/or GPU load.  
   2. High - enables high recognition accuracy, it requires less computing resources than for maximum accuracy.
   3. Fast - enables high recognition speed, but the accuracy becomes worse.
4. Select the computing resources use mode in the Strategy (3) drop-down list. The following options are available:
   1. Process  - mild mode: no more than 1 core for 1 license plate.
   2. System - default mode: all available computing cores are in use;
   3. Core  - strict mode: 1 core per stream.
5. From the Platform drop-down list (4) select the device on which the module will operate:
   1. Default — CPU resources are used for computing.
   2. TensorRT — NVIDIA GPU resources are used for computing.
   3. SV22, NCNN, HISI, Qualcomm, TFLite, OpenVino - other device resources are used for computing, if available. These options apply to platforms other than x86/x64.
6. Specify the minimum time interval lasting between the frames processing for the Interval (45) parameter (i.e. all frames within this interval won't be processed).

7. Select the profile of country license plates that are to be recognized in the Profile (56) drop-down list. 

   Note
   title Important!
   The profile of country license plates should match the selected country for plate recognition (see Selecting the сountry and license plate recognition SDK).

   The following options are available:

   • 0 - Common. This This profile must should be selected if By default country is selected.

   • 3, 6 and 9 – for more information see The LPR IntelliVision module functionality section.

     Info
     title Note
     The "9" profile enables recognizing correctly USA license plates with vertical letters on the plates. It can also be recommended to be in use when maximum recognition accuracy is needed if there is no high load on hardware.   Greater profile number enables more accurate LPR comparing to the solution with smaller number, but it is more demanding to computing resources.

     
     

8. Click Apply(67).

Configuration of the LPR IntelliVision module is completed.