Aria Insights - Saving Lives with Drones and AI Vision

  • Developed a super-resolution algorithm to enhance drone video footage in low lighting conditions and hazy environments
  • Implemented machine vision solution with object detection to identify defects, wear and tear, and deformities industrial containers
  • Created a platform for streaming and classifying drone footage aid inspectors during and after inspections
Company Overview
Aria Insights manufactures commercial drones for use in the oil & gas, defense, public safety, and telecommunications industries. Aria’s drones are designed to be flown in harsh conditions for durations of several weeks. Among other uses of drones including aerial deliveries, Aria Insights uses these drones for inspections of oil rigs, pipelines, refineries, and chemical tanks.

Problem
Performing inspections inside chemical tanks traditionally requires humans to enter dangerous environments. Humans are exposed to fall risks, toxic chemicals, and confined in low visibility spaces. Using drones for inspections removes the risk to human health and decreases inspection costs by circumventing the need for expensive scaffolding. However, drone inspections introduce new challenges, including:

  • Turbulence from drone flights and lighting conditions results in “shaky” and “grainy” videos.
  • Video quality and resolution is limited due to bandwidth limits of tethered drones.
  • Inspectors spend hours reviewing “useless” videos where drones are traveling between locations or inspection areas.

Solution
To overcome these issues and to maximize the value of the footage taken during inspections, our data scientists developed techniques to:

  • Automatically tag defects or features of interest in the drone flight videos using deep-learning based computer vision approaches. These approaches were able to identify faults in industrial containers including rust and cracks, as well as key areas within these containers such as nozzles.
  • Algorithmically improve the video footage by developing a bespoke super-resolution algorithm that exploits the interframe jitter from the turbulence experienced by the drone to increase the resolution of the images beyond the native resolution

Our platform engineers productionalized the data science models in a cloud-environment. The environment employs cloud-native design principles to scale depending on usage requirements. The environment:

  • Processes incoming videos is real-time and used the defect detection models to provide annotations in the videos that can be used by remote inspectors.
  • Indexes the videos to allow inspectors to fast forward or jump in time to parts of the flight videos containing areas of interest.

Impact
Aria Insights’ Drone Inspection solutions saves lives, time, and money. Human inspectors no longer need to enter dangerous tanks to conduct inspections, and inspections can be performed more efficiently with higher quality results:

  • Inspectors improve end-customer satisfaction and confidence by being able to create inspection reports with high-resolution images of defects and areas of interest.
  • Inspectors save time by reviewing footage that has been automatically analyzed and indexed, allowing them to jump between time points in the video with areas of interest.

We’d love to hear from you

Contact us for more information or to discuss what problem we may be able to help you solve.