Italian traffic camera and machine vision developer Tattile has released a compact axle recognition and counting system for high-speed tolling and other applications, based on deep learning AI (artificial intelligence) technologies.
Many toll systems around the world have made the number of axles of a vehicle a key parameter for determining the different tariffs applicable for certain types of vehicle. Even if a vehicle’s volume or dimensions are taken into account, this in itself does not unequivocally identify a vehicle class since two vehicles could have the same volume but a different number of axles. In addition, many cities are implementing congestion charging or clean air zones, with vehicle classification, often based on the number of axles, a key parameter in the application of restrictions. While intelligent traffic systems (ITS) have long been capable of performing tasks such as automatically reading and identifying license plates and measuring the speed of a passing vehicle, until now there has not been a smart solution for high-speed axle counting readily available, with many existing systems still reliant on the human interpretation of a picture taken at a tolling station.
As part of its long-term endeavours to simplify and raise the accuracy of ITS equipment, Tattile has now addressed this gap in the market and introduced its new system to completely automate the vehicle identification process through the use of AI-based image analysis. The company’s new Axle Counter automatically counts the number of axles on any vehicle driving at a speed up to 112mph (180km/h) on a highway. The entire system ensures correct axle counting and is mounted on a gantry above the highway. The infrastructure-mounted installation and stand-alone operational ability of the system means it is capable of analysing the side of all passing vehicles with its internal processing algorithms.
The Axle Counter operates completely automatically meaning that no post-processing of the pictures taken is necessary anymore, nor is there any more need for visual inspection and interpretation of the pictures by humans. Its high detection rate and high accuracy allow precise counting of the axles of any particular vehicle. The system includes two cameras per lane to ensure left and right sided analysis. An infrared (IR) illuminator allows operation both day and night. All image analysis operations, including the use of AI to automatically recognize the number of axles and wheels, as well as counting and classification, are performed on-board the system in real-time.
The Linux-based camera captures images with a high frame rate of up to 50fps and processes them with its internal deep learning algorithms that have been trained over a very large number of images and will continuously increase their knowledge base. The Axle Counter’s gantry installation has been made easy due to the Power-over-Ethernet (PoE) interface that provides a single cable connection to the camera for power and data transfer. The resulting metadata, together with the reconstructed image of the vehicle, provides supporting evidence to the road operator. For optimal performance, the Axle Counter can be triggered by different sources, allowing flexible interfacing with existing devices and integration with Tattile’s own ALPR devices.
