More information
Description
The project will develop a new generation autonomous data collection platform SMART FALCON. It is a powerful but highly miniaturized sensor control and operation system that can be mounted on a fast-moving or flying object (car, helicopter, drone). The operating system will be based on artificial intelligence that will analyze the collected data in real or near real time, identifying possible operator errors (e. g. the system moved too fast so the photo is out of focus), sensor errors (camera failed to activate in time), navigation errors (object is not in the photo), etc., and will ensure optimal quality control in real time.
During the experimental development activities, technology fusion and adaptation activities will be carried out, combining eSmart Systems artificial intelligence solutions with the engineering system developed by AISPECO and adapting these solutions to operate in real time. Tests, materials, component research, design development activities will be performed.
The activities of the Applicant and the Partner are directly complementary: the engineering equipment developed by the Applicant is intended for data collection; Partner''s software is intended for the processing and classification of this data. The synergy of systems, customers, markets and cooperation allows to maximize the potential of this project and ensure the growth potential of both AISPECO and eSmart Systems.
Summary of project results
One key challenge of transitioning to virtual overhead line inspections is the quality and consistency of image capture. There are many factors that impact this process – from camera quality, to angle of image capture to flight patterns when utilizing flight or drones. In real-life conditions, these factors often lead to subpar image capture, and issues such as improper exposure, blur and framing are commonplace. Sometimes operators are even forced to re-fly the lines to get useful data for virtual inspections, leading to inefficiencies and high costs.
Smart Falcon developed a solution that is able to identify in real-time when a target structure or part of the structure is in-frame, and automatically trigger a camera shot. The solution evaluates in real-time the quality of the image to determine if it meets the standards required, and otherwise another image is taken automatically without the need to circle back.
Edge computing algorithms for the improvement and automation of the image capture process for overhead line inspections were developed. This helps utilities capture images of their assets in a more efficient and reliable manner, with the goal of reducing costs associated with image capture by ensuring the right images are captured the first time. High-quality, consistent images also contribute to a significantly more efficient AI-assisted inspection process.
Summary of bilateral results
Partners in Norway asked a number of important questions that helped to pre-empt problems. For example, quickly identified a possible problem where an AI photographing an electric pole would focus not on a pole, but on the ground behind it – such a photo would be unideal. In addition, the partners had a lot of knowledge about AI, so we can purposefully recruit and configure AI faster