Technical Application Note 0002

Optimize Image Acquisition for Carrida

Revision: 1.0
Date: 2018-02-05
Contact: alpr.support@vision-components.com
Copyright: 1996-2018 Carrida Technologies GmbH, Ettlingen, Germany
Author: VC Support

Home page

Table of Contents

1   Introduction

This document describes how image acquisition should be setup in order to achieve good reading results with the Carrida ALPR engine. It is relevant for all Carrida users who want to deploy and use cameras in the field.

For the best possible reading results, images taken with a camera (IP cameras, industrial cameras, or the Carrida Pro camera) need to adhere to certain minimum standards. Achieving a reasonable image quality is sometimes not trivial, and a careful consideration of an installation must include

  • illumination conditions at the location
  • required sensor resolution
  • imaging geometry in general (camera position)
  • viewing angles
  • exposure times
  • gain settings
  • aperture settings of the lenses used

This document explains those factors and provides useful tips on how to setup a camera so that Carrida can achieve the best possible reading results.

1.1   Imaging geometry

The mounting position and viewing anle of your camera define the imaging geometry relative to the expected location of license plates.

Geometric factors which influence the image of the license plates are:

  • distance to the expected plate location
  • focal length of the lens
  • side viewing angle
  • vertical viewing angle

The following figure depicts those parameters graphically. The figure defines also our recommended maximum values for the viewing angles. The aspect angle of the camera relative to the car backside or front side (sideways viewing angle) shall not exceed 30 deg left or right. It will be possible to read at greater viewing angles, but expect to get worse reading results in this case.

im1

From the plate middle point to the camera angle should not accede 30° in any direction.

Vertical viewing limits should be in the range of 30 deg below and not more than 30 degrees above the license plate.

The horizontal viewing limits, which should be in the range of +/-30 degrees around the center of the license plate.

The projection of a license plate into the image creates also a rotation in the image plane, as shown in the following image. This rotation should not exceed more than 20 degrees relative to the x axis (the horizontal axis) of the image.

./images/imageTan1.png

Definition of the in-plane image rotation. The angle of a license plate should not exceed 20 degrees in either angle relative to the horizontal image axis. In the image above, the rotation is approximately 20 degrees.

Aside from the viewing angles, the combination of

  • sensor resolution
  • distance
  • and focal length

of the imaging system defines the size of the license plate in an image.

We recommend to setup a camera so that the characters of a license plate have a minimum size of 12 px. Carrida can read character sizes down to 8 px well, but this depends on the overall image quality. In addition, the overall size of a LP resolution should allow a gap of at least 1 px between adjacent characters in the license plate.

The maximum height of characters in an image should not exceed 150 pixels. Bigger character sizes can be configured into Carrida, but is generally not recommended.

In practice a typical setup would image the width of a car at about 300 pixels (independent of the camera resolution!). For example, for images with VGA resolution (640x480 px) the width of a car should be at least around half the image width. The following figure gives an example for the aforementioned character size recommendations.

./images/imageTan11.png

Definition of character sizes and recommended limits for imaging.License plates larger than the recommended size can be read, but this should be configured explicitly in the .INI configuration file.

Note

Note Sign The projection of license plates into an image as a result of viewing angle and license plate location can create an artificial rotation!

1.2   The camera setup

Taking control of the camera respective the imaging sensor is most important when it comes to obtaining a good image quality day and night and under all possible illumination conditions.

A camera in an outdoor environment or close to that (e.g. the exit of a parking garage) is subject to harsh illumination from the Sun which might create reflections, or from headlights of cars which may directly beam into your camera which is mounted at the gate.

Under all conditions, images should be sharp, with minimal motion blur, and of proper exposure. The license plates should not at all be overexposed - if you have to make a choice, its better to have slightly darker images than too bright ones.

As a general recommendation, the following exposure (= shutter) times are advised:

  • a shutter time of 0.1-3 ms on highways, for high speed vehicles
  • a shutter time of 0.1-3 ms in urban areas, for medium speed vehicles
  • a shutter time of 1-20 ms in parking situations, for slow vehicles
im2 im3
It is better to produce slightly underexposed images then overexposed images.

1.2.1   How to estimate the maximum allowed exposure time

The maximum possible exposure time can be calculated from the viewing geometry and image resolution. The basic principle is that a vehicle should not move more than 1 px in the image during the exposure time at at the maximum possible speed.

It is possible to either compute this value exactly using measurements of the visible street section, viewing angles, etc. or it can be approximated. By measuring the time T it takes a vehicle from top to bottom,or left to right of an image - whatever time is shorter - and the image resolution in this direction (i.e. number of pixels), the maximum allowed exposure time in ms can be computed as follows:

exposureTime(ms) = 1000*(T/resolution)

Hint

Hint Sign An average image brightness between 120-180 gray values is a good starting point for the overall image brightness.

The last important factor which affects image quality is the gain setting on your camera. The gain affects he sensitivity of your camera and it may be dynamic or not, depending on your camera brand and model.

The higher the gain, the grainier the images becomes, and ALPR will become harder. The limits of gain depends on your device, reasonable upper limits are typically 6-10 db.

1.2.2   Limiting exposure time and camera analog gain

A typical application for ALPR will limit the allowed exposure time and gain setting on the camera so as to not create blurred or noisy images. This can be achieved by enforcing a shutter and gain limit in your camera configuration either programmatically from your camera driver, or in the camera configuration dialog if the camera manufacturer provides a software for this purpose.

For example Point Grey sensors can be limited in their exposure settings using either driver functions or the Point Grey Configuration dialog.

The VC |car| Pro camera offers a setting in the configurator program to limit shutter and gain in auto exposure mode. The configurator allows you to set the shutter time and gain completely manually.

For many IP camera models it is not possible to directly control shutter and gain limits for proper automatic exposure.Some camera models allow parameter settings over the HTTP protocol, but this process is slow and not suitable for proper control in many application scenarios.

1.3   Using external IR illumination

To achieve the best possible image quality, we recommend to use external IR illumination in combination with IR bandpass filters on the camera.

A proven setup which is recommended by VC uses IR LEDS at a wavelength of 850 nm and a matching bandpass filter on the camera objective, which could either have a wavelength window of 720-850 nm or could be centered around 850 nm. Using the wider bandpass filter results in better imaging of the environment, a narrow bandpass filter results in much better contrast of the LP relative to the background.

The external illumination combined with the filter reduces the effects of glare from the sun or headlights from cars very efficiently. In combination with a wider filter options still allows daylight to pass so that the environment can be imaged as well.