Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Note
titleAttention!

To increase the performance of license plate recognition, you can use computing resources of the graphics card. To do this, when configuring the LPR channel object (see Selecting a device for license plate recognition), from the Use GPU drop-down list, select the required device. By default, only CPU computing resources are used.

The start (initialization) time of the license plate recognition algorithm using a graphics card can take several minutes. No license plate recognition events are received until the initialization is complete.

GPU files are included in the distribution package starting with Auto PSIM version 1.0.1.293. No additional installation is required.

For Auto PSIM earlier than version 1.0.1.293, you must activate computing resources of a graphics card:

  1. Download the distribution package to activate the GPU for the RR module from the AxxonSoft website.
  2. Unzip the downloaded archive.
  3. Put all files of the downloaded archive into the folder on the computer with installed Auto PSIM at C:\Program Files (x86)\Axxon PSIM\Modules64\UrmLpr\RR_gpu.

To configure the RR configure the RoadAR software module, do the following:

  1. Depending on the license type, go to the settings panel of the RoadAR RR module (fast) or RoadAR RR module (slow) object, which is created on the basis of the LPR channel object.
    Image Modified
  2. In the Number of streams field (1), enter the number of streams that the RoadAR SDK will use in parallel computations. A value of 0 disables the calculating process paralleling for the SDK of the RR recognizer uses during parallel computing. The 0 value (default) disables parallelization during license plate recognition.
    Note
    titleAttention!

    Increasing the number of streams during parallel computing also increases the load on the CPU. All changes are applied only after you restart Axxon PSIM.

  3. In the Maximum track size (frames) field (2), enter the maximum number of frames for one track that will be processed by the RoadAR RR SDK (the smaller the value, the faster the recognition result will be are given, but it will is also be less accurate). The default value is 0. The number of frames is not limited.

    Note
    titleAttention!

    If the LP recognition upon request is used (see Setting up the Camera of recognition upon request module), the Maximum track size (frames) should be must be equal to 1.

  4. Set the Skip unrecognized LPs checkbox (3) to ignore the vehicles with unrecognized license plates.Set the Use GPU checkbox (4) if it is necessary to use the NVIDIA GPU resources to increase the license plate recognition performance.
  5. By default, only the CPU resources are used.
    Note
    titleAttention!

    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.

    By default, only numbers license plates of the main country are recognized, which is specified in the settings panel of the LPR channel object  object (see see Selecting the country and license plate recognition SDK). If it is also necessary to recognize the numbers license plates of additional countries, click the  Select country countries button (54) and in the the Country selection window  window that opens, select the required countries.
  6. If it is necessary to determine the speed by video, specify the appropriate camera settings (65). The specified parameters should must correspond to the actually used camera and its mounting parameters.
  7. To save the settings, click the Apply button (76).

The configuration of the RoadAR software module is complete.

Info
titleNote

If several Roadar you create several RR recognition channels are created, the load on the CPU cores may can be uneven, because the even distribution mode isn't enabled by default. The enable it, set the 1 value for the string parameter of the IsProcessObject registry key (see Vertical solutions).

Configuration of the RR software module is complete.