Machine vision camera selection for ANPR

The selection of the right camera for an ANPR system requires careful consideration of several factors to ensure optimal performance and accurate license plate capture. Our recommendation for this project was our MER2-160-227U3M, based on the specifications explained as follows.
The researcher needed to use the camera with an Android law enforcement tablet, which didn’t have an Ethernet connection and would be placed next to the camera inside the Police car (thus, the distance was shorter than 4.6m) which led to using USB3.0 interface.

Resolution plays a critical role, as the system’s accuracy is essential for capturing clear and detailed images of license plates, especially under challenging lighting conditions. This ensures accurate character recognition even on smaller plates. For this ANPR project, the smallest characters to read were 1.5cm wide, which needed to be recognized in a 2.15m x 1.6m FOV. Therefore, this camera’s 1440px x 1080px were able to build a robust system.

Frame rate is another crucial factor. It determines the number of images captured per second, and a higher frame rate is vital for capturing fast-moving vehicles and guaranteeing accurate plate recognition. Aiming for cameras with a frame rate of at least 30 frames per second (FPS) is recommended to account for variations in vehicle speeds and ensure consistent capture. In this case, we had a high frame rate of 227fps at maximum resolution.

Pixel size also plays a significant role in image quality, particularly in low-light environments. Cameras with a larger pixel size (3.45um in this case), generally capture more light, resulting in improved image quality. This is crucial for capturing clear license plates in diverse lighting conditions, ensuring the system's effectiveness throughout the day and night. Furthermore, for the same reason, the monochrome version was preferred, as it’s 3x more light-sensitive than color cameras.

Finally, the choice between global shutter and rolling shutter cameras is crucial. Global shutter cameras capture the entire image at once, minimizing distortion caused by moving objects, which is essential for capturing clear license plates of vehicles in motion. While rolling shutter cameras may be cheaper, for ANPR applications, global shutter cameras are highly recommended for their superior image quality, especially when dealing with fast-moving vehicles.

Selecting the Optimal Lens

The researcher also needed assistance in selecting the optimal lens for their ANPR system. Ensuring clear capture of license plates, especially for vehicles at varying distances (5 to 15m in this case), was crucial.

To achieve this, we followed a similar approach as used in previous projects. We focused on two key factors: field of view (FOV) and working distance (WD). The FOV needed to be wide enough to capture the road (2.15m) while maintaining a suitable WD for clear focus.

Through discussions and calculations (using our lenscalculator), we determined that a 16mm focal length lens would be ideal. This lens offered the necessary balance, providing a wide enough FOV for capturing license plates at the expected distances while upholding a sufficient WD to avoid focusing issues.

Furthermore, we prioritized minimizing lens distortion, ensuring accurate license plate recognition.  Distortion can cause characters to appear warped, hindering identification.  Similar to previous projects, we recommended a lens with minimal distortion (typically below 1%) to guarantee clear and undistorted images, crucial for effective ANPR functionality.

By applying this methodology, we successfully assisted our client in selecting the optimal camera-lens setup for their specific ANPR system, ensuring clear and accurate license plate capture, which they keep using to this day.

Lighting for ANPR

As mentioned, this researcher needed to capture license plates clearly, no matter the time of day. We understood their need, as clear images are vital for accurate license plate recognition.

Similar to other projects, we focused on optimizing lighting to enhance image quality. Here, the "object" of focus was the license plate.

Unlike photography, capturing license plates requires different lighting approaches. We recommended using Near-Infrared (NIR) spotlights. Invisible to the naked eye, these spotlights operate in the same range as the client's camera. They act like invisible flashlights, illuminating the scene without disrupting the night environment. With NIR spotlights, the ANPR system can still see license plates clearly even in low-light conditions, since this camera works in the NIR range.

Finally, we addressed the issue of reflections. Depending on the material of the license plate and the camera angle, light can bounce off and create glare. To minimize this, especially during the day, we suggested using polarizing filters. These filters act like sunglasses for the camera, blocking specific light waves that cause reflections. This keeps the image clear and helps the ANPR system read the license plate accurately.

By considering both NIR spotlights for night and polarizing filters for daytime, we helped the client achieve optimal illumination for their ANPR system, ensuring clear license plate capture day and night.

Image processing software for ANPR

A camera, lens, and lighting are just the first pieces of the puzzle when it comes to building a complete machine vision system. For tasks like ANPR, computer vision software is needed to analyze the captured images and extract meaningful information.

Fortunately, our systems offer flexibility when it comes to software choices. They're GenIcam compliant, meaning they work seamlessly with various third-party software options like MvTec Halcon, NI LabVIEW, Cognex Vision Pro, and even popular platforms like MATLAB and OpenCV.

We offer a free Software Development Kit (SDK). This handy toolkit helps you capture images and control camera parameters, making it a great starting point. The SDK is compatible with various platforms, from regular PCs to industrial setups with NVIDIA TX series or Raspberry Pi devices. It even supports diverse operating systems like Windows, Linux, and Android. There are plenty of Open-Source programs built for ANPR that work with our SDK.

This researcher built their program, together with our SDK, to recognize license plates. This program also made it possible to change automatically the camera’s parameters according to the environmental conditions.

For projects looking for a whole Software solution pack, we recommended Zebra Aurora Vision Studio. This powerful yet intuitive software offers a visual interface that resembles a toolbox, making it easy to learn and use. With Aurora Vision, users can recognize the number plates. The free Lite version, equipped with all standard algorithms, served as a great starting point for the client to explore the software's capabilities. 

By offering various software options and user-friendly tools, we empower customers to choose the solution that best fits their expertise and project requirements, ultimately leading to successful machine vision implementations.

ANPR applications

Automatic number plate recognition ANPR has transcended its security checkpoint origins. Today, it streamlines operations and enhances security across industries:

Traffic Flow: ANPR automates toll collection, analyzes traffic patterns, and identifies speeding vehicles, improving flow and generating revenue for safer roads.

Parking Management: It facilitates authorized entry/exit and monitors unauthorized vehicles in parking lots, ensuring convenience for legitimate users and revenue for operators.

Law Enforcement: ANPR assists in recovering stolen vehicles, tracking missing individuals, and apprehending criminals, contributing to community safety.

Fleet Management: Companies use it for real-time vehicle tracking, route optimization, and driver behaviour monitoring, leading to cost savings and efficiency.

Border Security: ANPR strengthens border security by monitoring vehicles, identifying those of interest, and verifying travel authorizations.

This versatile technology reads and interprets license plates, fostering smoother operations, enhanced security, and convenience across diverse applications.

Support for ANPR?

Would you like to have support from one of our machine vision experts for creating your own Automatic Number Plate Recognition (ANPR) system? Or another similar vision system? Don’t hesitate to reach out by using the form below!