ANPR, ALPR, License plate recognition, Number plate recognition, smart camera, high performance computing, parallel computing, computer vision, machine vision, artificial vision, FPGA, image processing, embedded systems, traffic enforcement, dangerous goods, ITS, intelligent transport systems, travel time, average speed enforcement, free flow tolling ANPR
ANPR, ALPR, License plate recognition, Number plate recognition, smart camera, high performance computing, parallel computing, computer vision, machine vision, artificial vision, FPGA, image processing, embedded systems, traffic enforcement, dangerous goods, ITS, intelligent transport systems, travel time, average speed enforcement, free flow tolling ANPR ANPR, ALPR, License plate recognition, Number plate recognition, smart camera, high performance computing, parallel computing, computer vision, machine vision, artificial vision, FPGA, image processing, embedded systems, traffic enforcement, dangerous goods, ITS, intelligent transport systems, travel time, average speed enforcement, free flow tolling ANPR ANPR, ALPR, License plate recognition, Number plate recognition, smart camera, high performance computing, parallel computing, computer vision, machine vision, artificial vision, FPGA, image processing, embedded systems, traffic enforcement, dangerous goods, ITS, intelligent transport systems, travel time, average speed enforcement, free flow tolling ANPR
fAQ
 
ANPR, ALPR, License plate recognition, Number plate recognition, smart camera, high performance computing, parallel computing, computer vision, machine vision, artificial vision, FPGA, image processing, embedded systems, traffic enforcement, dangerous goods, ITS, intelligent transport systems, travel time, average speed enforcement, free flow tolling ANPR

     

 

This is a collection of common questions asked to Imagsa regarding its artificial vision cameras and technology.

Imagsa's value proposition in the ITS sector

250 fps capture and analysis in Real-time, to achieve best ALPR performance

The importance of using a megapixel image sensor

The negative effects of using a 752x480 pixel camera

Do not lose resolution as the vehicle travels

Multiexposition to overcome challenging environmental conditions

Motion detection versus License plate detection

Minimise Bandwidth usage

Temporal precision of every license plate detection

The importance of an easy and quick installation

How and where to use Atalaya ALPR cameras

Imaga's value proposition in the ITS sector

In the ITS sector, Imagsa Technologies aims to provide smart cameras with the following added value:

  • Full traffic analysis: Detect and recognise all and every type of vehicle, at any speed and traffic condition.
  • Anytime, Anywhere: Perform under any environmental condition, 24x365 and in every country.
  • Versatility: Become a unique platform to integrate many different road traffic control applications.
  • Cost Effective: Develop traffic control and management systems in a low cost of ownership.

250 fps image capture and analysis in real-time, to achieve best ALPR performance

Imagsa Technologies has developed an autonomous (no external trigger), versatile and highly-precise license plate detection technique. A demo of this technique is available online.

The aim behind the license plate detection technique developed by Imagsa is to obtain the best quality image of the vehicles' license plates in terms of resolution and sensitivity. In order to achieve this, it is mandatory to control the capture of images from the camera. Imagsa's ALPR cameras effectively do this by capturing images at 250 fps, varying some capture variables, and it analyses hundreds of images in real time. The outcome is the best quality image of every license plate independently of environmental or traffic conditions, so the OCR software can reach its best performance (above 90% recognition rate).

The importance of using a megapixel image sensor

In an European standard license plate, with a character stroke of 1cm, the OCR providers suggest a minimum of 3 pixels per 1 cm to achieve high reliability rates (above 95%). Therefore, to fully cover 4 meters of the road (3,5 meters lane width plus 0,5m overlapping), an image sensor with resolution of 400cm x 3pixel/cm = 1.200 pixel is needed.

Additionally, vehicles changing lanes can be detected if a megapixel sensor that fully covers 400 cm of the road (3,5 meters lane width plus 0,5m overlapping) is used.

The negative effects of using a 752x480 pixel camera

When a 752x480 image sensor is used, the following take place:

A)  To cover 4 meters of the road, the resolution of the license plate image will be: 752pixel / 400cm = 1,88 pixel/cm. With this resolution it will not be possible to reach a decent license plate recognition rate (OCR software providers suggest to use a minimum resolution of 3 pix/cm to achieve 95% recognition rates).

B)  To achieve a 3 pix/cm resolution with an image sensor of 752x480 pixels, the lane width that is covered is: 752pix / 3pix/cm = 250 cm. Therefore, only the central part of the lane will be covered with the target resolution of 3 pix/cm.

C)  License plate detection and recognition can underperfom when a 752x480 image sensor is used: issues with the installation of the camera (out of focus, dust in the lense, et.), and also sensitivity to harsh metereologic conditions (rain, snow, fog, etc.) can affect the quality of the image and the performance of the ALPR system.

Do not lose resolution as the vehicle travels

Capturing the vehicle and an image of the license plate in an exact and deterministic point of the road is essential to guarantee the target resolution of 3 pixels/cm, even when a megapixel sensor is used (as explained above).

Imagsa's ALPR cameras define a virtual loop in the lowest part of the overview image (the red rectangle: 1,5m), as it is the closest point to the camera. In this region, hundreds of images are captured and analysed in real-time, to detect the license plate of the vehicles at the target resolution of 3 pix/cm. This approach ensures that the OCR software achieves its highest recognition rates.

Multiexposition to overcome challenging environmental conditions

The following images are captured by the same camera at the same point and same moment (seconds of difference). It can be clearly observed that the license plates have different illumination conditions: overexposition, shadow from the vehicle, etc. These images prove that the illumination change locally. Therefore, the correct exposition time is different for each license plate of every vehicle, and it depends upon the sun reflectivity on the license plate, the license plate dirtiness, shadows, etc.

The previous example shows that an ALPR camera has to:

  • Adjust the exposition time in real-time to overcome the sudden illumination variations in the scene (environmental illumination),
  • And, perform multiple captures of the same vehicle with different exposition times, to overcome the local illumination variations affecting the license plate.

Imagsa's ALPR camera captures multiple images varying in real-time the exposition time depending on the average scene illumination, position-orientation of the camera towards the sun, reflectance of the target. Imagsa has filed a patent on the estimation of sun reflectance into the license plate using a GPS device embedded in the smart-camera.

Motion detection versus License plate detection

Standard ALPR cameras with video triggering detect a vehicle by means of motion detection techniques. Motion detection consists on analysing the differences between video frames. When a difference occurs, this approach considers that there is a vehicle and it captures an image to read the license plate. This technique has one main drawback: it has false detections caused by shadows of a vehicle travelling in the adjacent lane, or even by shadows of trees next to a road.

 

Imagsa's ALPR cameras implement a license plate detection technique. Basically, it consists on continuously capturing and analysing images to detect a license plate pattern. When there is a license plate pattern it means that a vehicle is in the field. This approach avoids the false detections that are common in an ALPR camera based on motion detection.

Minimise Bandwidth usage

Imagsa's ALPR cameras minimise the network bandwidth usage by:

1) Segmenting the vehicle license plate from the overview image. The high-resolution image of the license plate has a size of 4kBytes.

2) If an overview image of the vehicle is required by the application, a JPEG compressed file can also be delivered.

Temporal precision on every license plate caption

Journey Time Monitoring applications require that the ALPR cameras capture images with a precise time stamp. Imagsa has implemented a GPS based caption system in the Atalaya ALPR smart cameras. By using the GPS, each of the 1.000 images per second are marked with a precise time stamp delivered by the GPS. The precision is below 1ms.

Also, all the Atalaya ALPR cameras installed in the road are synchronised by means of the universal GPS system. Therefore, a highly precise and versatile (high-speed vehicles in short trams) journey time system can be developed using Atalaya ALPR cameras.

The importance of an easy and quick installation

In addition to performing an accurate license plate detection and recognition, an ALPR camera needs to be easily and quickly installed by road operators.

The ALPR camera Atalayais delivered with mounting accessories and software tools that ease the installation and parameterization of the camera in different types of installations (in a gantry, in a roadside pole, etc.).

How and where to use Atalaya ALPR cameras

Along with the delivery of reliable and versatile ALPR cameras, Imagsa Technologies provides its experience in the ITS sector to provide advice on how to use the technology and cameras. Indeed, Atalaya ALPR cameras are being used in numerous traffic control applications such as: Journey Time Monitoring, Variable Speed Schemes, Dangerous Goods Transportation Control, Average Speed Enforcement, O.D. patterns, vehicles' black lists, Counting and Classification projects, and many other urban and interurban projects.

Please, submit your questions here.