Powered by computer vision technology to recognise different types of road users, including pedestrians, cyclists and motorised vehicles, within the sensor field of view and track them from frame-to-frame. Over 150 transport authorities have used our multimodal traffic data to understand road user behaviour and support infrastructure changes.
Smart Temporary Surveys make remote checking possible, which would be impossible with existing technology from a single sensor, and provides tools to ensure that the survey is delivered to the required specification.
As a market leading, UK made AI-powered technology, our sensors independently verified accuracy levels are unsurpassed in the industry and are continually improving. This significant advantage against our competitors is due to the nature of Artificial Intelligence and the fact that we have over 5,500 permanent traffic sensors deployed and have counted 40,000,000,000 road user behaviours.
VivaCity’s sensors are able to count and classify the following modal classes bidirectionally as standard: pedestrian, cyclists, motorbikes, cars, taxis (black cabs only), LGVs, OGV1 (rigid HGV with up to 3 axles), OGV2 (articulated HGV or HGV with 4 or more axles) and buses.
Our sensors can monitor bidirectional flows at each requested location. Each device can provide multiple count-lines, enabling a single sensor to provide multiple data feeds e.g. one count line for the pavement, one for the segregated cycle lane and one for the main road carriage way.
In addition to the link count style surveys that can be provided using the above, the same technology can be used to generate junction turning counts for complex intersections, and path/desire lines, to visualise road user trajectory and interaction with one another in new ways. Furthermore, additional features, such as classified speeds, journey times and near miss data, are in development and will be possible to purchase (at an additional cost).
The SIM cards within our sensors enable 3G/4G connectivity and remote access is available to all sensors. This enables a full, real-time view of sensor up- time, remote debugging and maintenance, and remote updates.
Due to the depth of our training data sets, our system is able to maintain high accuracy during low light or in poor weather conditions including fog, rain, snow and sun glare (see below). Count accuracy is unaffected at night time provided the road has street lighting.
Method of detection
VivaCity sensors use state-of-the-art Computer Vision algorithms for the detection, classification and movement of road users within the field of view. The approach is a variant of supervised learning, where the algorithm is trained on a dataset of thousands of annotated images (with millions of real-life images available depending on the class). The AI is applied to each frame of video, typically at 10 frames per second or higher. Built with privacy at its core, the images captured are blurred on edge and no personal information such as faces or number plates are identifiable, making the system GDPR compliant. It is this approach, building proprietary algorithms, which has established VivaCity as the class leader in accuracy and reliability, and sets us apart from alternative out-of-the-box computer vision systems.
Sensor hardware specification
Weight and windage: Sensor and battery package are lightweight and can be installed by a single Technician in under 5 minutes, with no power tools required. Windage on the column is less than half of a “type A” small road sign.
Edge processing: VivaCity was the first company in the world to deploy a fully-integrated, GPU edge- based processing camera based on Nvidia Jetson chip architecture as early as 2015. Our on-board edge processing capability ensures that data can be processed in real-time and provided to you in near- real time. This on-board processing also ensures that sensors only transmit anonymised GDPR data from our sensors to clients.
Processor: The processor in our sensors has been specified to process video at a rate of 15 frames per second in areas across up to 4 lanes. Processing speed can be optimised for different use cases.
Lens: Wide angle view to capture local area data. Height: Mounted at a height between 4m and 8m. Our sensors are easily and quickly mounted onto existing street furniture such as street lighting columns and traffic lights (via a signal extender pole) using a stainless steel band attaching the sensor bracket to the column. No ground works are required.
Power: Batteries can provide power for up to 48 hours. They are “hot-swappable”, meaning surveys can be performed with continuous counts for up to 30 days.
System accuracy: We have extensively benchmarked our software performance, and currently achieve a 98+% count and classification accuracy for all classes during the day. Count accuracy is unaffected at night time provided the road has street lighting.
