The hype around drones, their prowess and possibilities, their supernatural abilities started becoming the rage a decade ago. And then reality set in.
Governments and regulatory bodies are still trying to figure out how much they want to regulate commercial drones. Skilled professionals to operate drones are still hard to find. The world is still waiting for other technologies such as IoT, sensor technology, telemetry data processing speeds, computer vision, and such to reach a certain level of sophistication.
And so on and so forth!
Yet, we have now helped two Fortune 500 companies productize and find phenomenal use for drone technology. From Wizard of Oz testing to validate data to building the software architecture that delivers true benefits — we’ve done the gamut.
For an older HealthTech initiative, we enabled one of the earliest successful human organ deliveries by an unmanned aerial vehicle (UAV). The other is the true hero of this blog post: drones and their application in the oil and gas industry.
The global drone surveillance market for energy and power is projected to grow at a CAGR 18.9%, to $476.5 million during 2021- 2028. And we couldn’t be more excited about that prospect.
Monitoring, preventing, and managing emissions have become priority number one for governments and large corporations everywhere. One Fortune 200 company in the space worked with Zemoso to build and productize an aerial system that helped detect and prevent leaks at remote plants: oil wells, subsea stations, oil rigs, and more.
So, how does aerial drone-based monitoring work: data across different wireless sensors and drones are collated, processed, and analyzed in real time to identify if there is a leak, what is the rate of that leak, and notify the stakeholders to act as needed.
Zemoso’s scrum teams worked as a part of a leading conglomerate’s internal team to help develop the software product that would ultimately take advantage of all the data streaming in from drones and sensors, empower decision makers to take immediate action and bring about real reductions in emissions at remote energy plants.
We aligned with our stakeholders on the minimum feature set and created the blueprint for this complex product. After that, we ran weekly sprints and scrums to design, build, and deploy the product within eight months.
With regulations and traffic not being a major point of concern in remote oil and gas production plants, we set off to deal with the other engineering challenges that would determine the success and failure of this fantastic product we were co-building with our enterprise customer.
On the engineering front, we stuck to all the agile best practices: using microservices architecture, service choreography, GraphQL layer for faster deployments, CI/CD, automated testing, containerization, and so on. The two main challenges we focused on were:
Firstly, to expedite deployment, we decided to use a third-party API to gather and transmit geographic information, making it easier for operators to act. This helped us gain access to superior geolocation tech without slowing down the actual build timeline.
In addition to automating testing in general, we used Selenium to ensure accuracy and speed. We used Postman to automate API testing at scale.
We used Cloudfront content delivery network (CDN) to deliver data, and videos fast, with low latency. We also used AWS Network Load Balancer to prepare for sudden shifts in data traffic; this ensured that the solution could handle millions of requests per second.
We used PostgreSQL with Redis for cache, ensuring sub-millisecond response times, enabling millions of requests per second. Amazon S3 was used to store generated reports and images, making the product even more robust in handling data in multiple formats.
We used Python, MongoDB, and PyMongo for analytics, and Pytorch for computer vision apps to process the visual data (images and videos) coming in from the drone.
Images and videos collected via drones, processed using AI and computer vision helped generate the right notifications. These images and videos are ingested, sorted, and analyzed to generate insights in real time. Region-based Convolutional Neural Networks (R-CNN) helped detect objects in any image.
P.S. Since we work on early-stage products, many of them in stealth mode, we have strict Non-disclosure agreements (NDAs). The data, insights, and capabilities discussed in this blog have been anonymized to protect our client’s identity and don’t include any proprietary information.
