We’re constantly looking to innovate and to improve our technology; specifically to push the boundaries of DAS. After all, pioneering sensing technology is what our customers expect from us.
Steve Cammish, our CTO, recently presented at the 2022 NVIDIA GTC conference for AI developers explaining how we do just that. He focused on how we are overcoming some of the challenges of processing exceptionally high volumes of data. He also outlined how machine learning and edge-AI quickly improves the accuracy of our devices to provide extremely high confidence in alarms.
Our DAS technology needs advanced algorithms and processing techniques to analyse changes in light patterns created when a fibre undergoes strain, and to identify and to categorise any disturbance. That’s where the latest computing platforms from NVIDIA come in.
The first challenge is processing up to 600MB of data per second - an exceptionally large volume of data! It would take too long to transfer that amount of data to the cloud for our customers to get the real-time insights they need and would also be too expensive. In order to detect with the sensitivity, and to cover the distance of fibre being monitored, we require intense edge computing.
The data is processed using the NVIDIA Jetson™ edge AI platform and NVIDIA CUDA®, which delivers the computing power for the detection of the different types of events our customers are interested in, such as mechanical strikes on the ground, leaks from pipelines, footfall or fence climbing. Once we know the events, we have much lower data rates and are then able either to process them on the device, or in the cloud, to create alarms.
Reducing false positives of alarms is the second challenge. It’s a critical demand from our customers as it means they can reduce wasted resources, for example sending out security teams to investigate incidents. They want a system that will produce an alarm, at the right time.
We work closely with our customers to really understand the conditions for their application. For example, between five and 10 spade strikes might be considered normal activity that doesn’t warrant investigation, but anything more than this does. Knowing this information means we can fine tune the device so we generate threat alarms that are correct.
The ultimate goal of our analytics is to get it right every time; new AI techniques help us to be even smarter. We have introduced new data flows for machine learning into our offering to improve detection accuracy.
We collect data, we label it, process it and create trained models that we test before we deploy to devices on live sites. Using NVIDIA TensorRT, a software development kit for high-performance deep learning inference, allows us to execute the models efficiently and to identify additional data we need to tune the models in a subsequent training cycle, for constant improvement.
NVIDIA TensorRT™, a software development kit for high-performance deep learning inference, enables us to execute AI models efficiently and identify additional data needed to tune the models in a subsequent training cycle for constant improvement.
To learn more about how we are using edge AI to create accurate insights from data, watch the on-demand session featuring Steve Cammish, our CTO, at NVIDIA GTC.
Find out details of our collaboration with NVIDIA.
Fotech appoints Jean-Yves Le Gall as Chief Sales Officer as the company continues to scale up
DAS photonics can’t solve the world’s energy crisis on their own. But they can play a crucial role in asset monitoring to conserve resources. Find out how.
Fotech is a preferred partner of NVIDIA, harnessing their advanced AI platforms to enhance our own DAS photonics sensing technology. Our CTO, Steve Cammish, recently presented at the 2022 NVIDIA GTC conference, read more here...
Fotech has appointed Jairo Linares as Regional Sales Manager for Latin America (LATAM).