The benefits of ROI explained

Less pixels to process

Next to a higher framerate adjusting the region of interest by cropping has another beneficial effect. Less pixels per image speeds up recognition. Less pixels mean less processing power is needed.

Less background noise

Besides that cropping minimizes the amount of random background that needs to be analyzed. To use the power of a Gatekeeper-X we only want to keep the pixels we need for recognition.  Make sure the cropped area covers the complete range of positions for capturing a license plate.

Speeding cars

A high framerate is desirable when recognizing speeding cars. The more frames per second the more frames are captured of one car, the better recognition works.

Higher threshold values

Thresholding is the tool to correct circumstances that influence a straight forward recognition like head lights shining in the camera, license plate that are difficult to read.

The threshold defines the amount of times a license plate needs to be recognized identically to become a valid Anpr result. An increase in frame rate allows for higher threshold values. The higher the threshold values the more accurate the result will be. The threshold and it’s parameters can be configured in the ANPRThresholdAxon.

1. Selecting the region of interest

The viewpanel of the IDSAxon/IDSPeakAxon diplays the live view of the IR sensor. Rubber banding can be used to select an area of interest. Hold left mouse button, draw the rectangle of the desired region of interest and let go. The view of the Axon will immediately change. To reverse the crop click 'reset image size'. To save the new settings right click the Cortex in the CortexNavigator and choose save cortex.

In previous version rubberbanding is not possible. Read instructions how to do it here.

Using the "crop map" to select the best possible ROI

A special feature to get the best possible recognition area is the "crop map" function of the IDSAxon/IDSPeakAxon.

The coordinates of recognized license plates are stored and are used to determine the best crop area.

The "crop map" is an overlay, that can be activated by right clicking the view of the IDSAxon/IDSPeakAxon and choose "crop map". The overlay will show the crop map as a cloud, that represents all past recognitions.

The cloud is the most dense (white) in the area most license plates are recognised.

In the black parts of the image, no license plate is ever read.

In the screenshot the crop can even less high.

2. Check license plate recognition in the GatekeeperAxon

The GatekeeperAxon viewpanel (right) will show the cropped image with an ANPR overlay. To enable the overlay, right click the image and choose 'anpr'. Read more about the overlay and the meaning of the numbers here.

When the overlay is not visible the license plate is not recognised. The license plate is either too large or too small (width should be 100-350 pixels).

Another reason could be, that the ANPRAxons are not configured accurately or the minimum license plate size is not set properly in the ANPRThresholdAxon.

3. Selecting the region of interest (previous versions of CortexClient)

The image size and position can be altered through the sliders in the IDS settings.

Position x:

X coordinate to determine the position to crop the image

Position y:

Y coordinate to determine the position to crop the image

Width:

The desired width of the image
Height:

The desired height of the image

Reset size:

To revert back to the original image

Set new size:
To save these settings

Important:

The framerate of the IDS camera can only increase when the height of the image is adjusted. Adjustment of the width will not effect fps. Cropping to increase framerate is done in the IDSAxon, not with the ImgCropAxon.

4. Some examples of cropped images