In order to rapidly respond to escalating risks, Valerann has created a Digital Twin of the road ecosystem. Using a unique combination of Object Detection, Object Tracking, and Trajectory Smoothing. Valerann is able to do things like model the position, speed and classification of every vehicle in real time. Having abstracted the key information from the noise, Valerann performs novel algorithms to detect the presence of events such as Stopped Vehicle, Animal on Road, and Blocked Lane.
Not content to dwell near the cutting edge of Artificial Intelligence research, Valerann is pushing its Machine Learning capacity to the next level by leveraging live weather data inferred directly from road-facing cameras along with Natural Language Processing of social media feeds. Armed with these powerful sources, we are able to predict when there is elevated risk of an accident due to inbound thunderstorms coinciding with the increased traffic following a football match, for example.
Valerann is able to cast light on the road through the use of our proprietary Real-Time Fusion Engine, offering traffic operators the most complete perspective of the current road conditions.
Nowadays, information about the condition of the road comes from all shapes and forms — cameras, weather reports, social media, loops, connected vehicles and Lidar. Inherently, any individual data source is always incomplete and has its limitations, uncertainties and blind spots. However, when combined, they tell a comprehensive story about what’s happening on the road. Data fusion allows road operators to leverage the best of everything. Using custom knowledge graphs and optimal estimation, we assimilate data, measurements, information and their associated uncertainties from a multitude of sources and derive real-time and actionable insights about the road and its users.
For example, by fusing weather reports, radar measurements, real-time video feed and journey time reports, we can anticipate the formation of a traffic jam; by fusing vision-based object detection, the behaviour of vehicles on the road, and historical accident records, we can infer the occurrence of an accident. Finally, our data fusion engine gets smarter over time as it learns the correlation amongst and intricacies of each data source.
Enter Valerann. Streaming all cameras at 30 frames per second straight into the Cloud, our algorithm detects the accident and automatically alerts the nearest Patrol Vehicle to respond as they see fit, such as by setting up a cordon or phoning the emergency services. What’s more, we are able to back-up all 2.5 billion frames of your data every day and can replay conditions from any historical date since the installation. Rather than allowing this archived data to gather dust, we continuously run pattern detection algorithms so as to alert Operators to present traffic anomalies or recognise conditions that increase the likelihood of imminent events.
Even operating with infallible equipment, the sheer volume of data renders many ostensibly simple tasks—such as moving a PTZ camera to focus on a localised accident—impractical to perform manually. Valerann aims to automate as many of these jobs as possible, liberating Operators to focus on the tasks that require human nous and industry expertise.
Our goal is to learn with you and scale with you, so you can set your horizons beyond the screens of your control room.