The overall vision of the ARIA Accelerator project is to prototype, pilot and commercialise our revolutionary bird detection (bird/non bird), tracking and species identification system for GW-scale wind farms that uses computer vision and AI. The patent-pending ARIA system applies a two-step process for birds’ classification: 1) a convolutional neural network (CNN) in the short range to be able to classify a bird from a 30x30 pixel image at <1km distance; 2) a combination of CNN to encode the visual information and a temporal neural network to process temporal information at >1km distance. ARIA has been in development for 2 years by AI and computer vision experts, as well as ornithologists, to achieve accuracies of ~95% detection at 2km and ~78% identification of 8 species (e.g. European Herring Gull) combining bird features (e.g. wing span, wing-beat frequency) and domain knowledge (e.g. weather correlation, migration patterns or seasonality). ARIA processes the images on-site to reduce latency and enable an accurate flight path prediction (a root mean square error (RMSE) of up to 49 metre of the ground truth) and estimation of collision risk.
Our technology has a major market potential compared to current state-of-the-art technologies. To increase our commercial success and impact, we have created a Strategic Business Plan, as a detailed roadmap, describing all necessary activities we need to undertake to ensure a successful market introduction. Now, we seek to develop our final Business Innovation Plan (BIP), to be created in this Accelerator project. As the envisaged Accelerator project builds upon a validated prototype technology, most of our efforts will be focused on piloting and commercialisation activities.