Alerting to live and unpowered wires and masts

How We Use AI

We use a type of AI called machine learning to create a neural network—the core component responsible for recognition of objects appearing in the video stream.

Training machine learning models in the lab

Data from real flights

Data from simulated flights

Training
dataset

Testing
dataset

Validation
dataset

AI component within the airborne application

Input

Video stream from cameras

Pre-processing

Preparing frames for analysis (using classical software)

Algorithm frozen after training and testing, not evolving or learning in flight

Convolutional Neural Network

Identifying span of wires and outline of masts

Post-processing

Sending output data to pilot display, ensuring continuity of identified objects

Output

Localization, distance, and time to impact