Data’s Journey — Part 3: Data Organization

Author
Justin Lee
8
min read
|
Sep 25, 2023
Data Organization
Detect, in partnership with Newfoundland & Labrador Hydro (NLH) and Stantec, is excited to introduce a new series: “Data’s Journey”. This is a five-part series that takes you behind the scenes of our critical inspection processes. We’ll provide you with a deep dive into each of our steps, starting from the initial planning phase to visualizing collected and organized data on our platform, SCOPE.

Overall Project Stats

  • Structures Inspected: 1314
  • Photos Taken: 39,401
  • Defects Detected (compared to previous inspection): 10X
  • Total Distance: 220 km
  • Total Inspection Time: 22 Days
  • In-field Team Composition: Pilot, Safety Officer, and Data Manager
    ‍
“Management of many is the same as management of few. It is a matter of organization.” — Sun Tzu


In this third installment, we’ll discuss data uploading and data organization. Given the 1314 structures and +39,000 photos, it is important to build robust and automated systems to handle this volume of data. This project involved two parallel transmission lines, L23(A) and L24(A), spanning approximately 220 km between Wabush and Churchill Falls, NL.
‍

Data Models and Entity Relationships

First and foremost, understanding our data practices is crucial, as they have empowered our team to conduct inspections both safely and efficiently. Data models serve as the foundational blueprints for our entire data management system, streamlining data organization and specifying relationships between distinct data elements. Entity-relationship diagrams (ERDs) further enhance our comprehension by visually mapping out these specific connections.

Imagine you are managing a city’s transportation system. You would need to manage roads, public transit, and bike paths ensuring each neighbourhood is accessible. You would need to make decisions on the best means of transportation across these different routes, understanding the changing needs of your city. Similarly, a data model looks at building a coherent structure, determining how data pieces fit together and relate to one another for efficient access and organization.

ERDs are akin to a detailed city map, showing how these unique modes of transportation come together. A small road may be connected to a major road, acting as the main artery through your city. Bike lanes, bus stops and train stations may intersect or be alongside this major road. These unique elements all interact with each other, with specific rules and frameworks to ensure safe and efficient modes of travel. An ERD looks to do the same thing with data, highlighting specific linkages between different data elements so that we can navigate our data infrastructure seamlessly. A well designed data model and ERD standardize how different data entities relate and interact with each other.

To delve deeper, consider a structure as an example: while multiple images can be associated with a single structure, and several images might reference multiple structural members, a structure is uniquely linked to just one powerline. Conversely, a powerline can encompass numerous structures.

It is important to understand the underlying piece as it serves as the foundation for our image pipelines and how it enables the later stages of the inspection. Data’s journey into our platform is managed using our data models.

Relationship between Structure, Images and Defects

The importance of metadata

Once the in-field team has collected the images, these are uploaded onto our platform. It is here where the data model and our pipelines come into play.

First, it is important to understand what data exists when images are taken. When an image is taken using our drones, the image of the structure is made readily available for our annotators to review. The challenge arises when you have +39000 photos, especially when there’s no clear understanding on their associated structure or viewing direction.

Metadata, in its simplest form, is data about data. Think of it like the label on a folder in a filing cabinet; it gives you essential information about what’s inside without having to open the folder itself. This information accompanies every image we take and becomes the key to addressing the challenge. It provides context and helps us organize our images, making the image metadata an indispensable part of our process.

Sample structure image

When images are taken, they are geotagged with the date and time it was taken. Furthermore, with our drones, the heading and angle of the photo alongside the elevation is available. What this allows us to do, using our proprietary software, is automatically associate these photos to any given structure quickly and accurately.

Basic metadata overview

In traditional drone-based inspection services, it’s not uncommon for teams to receive vast quantities of photos in an unorganized manner, necessitating manual sorting and organization. With our data models and software, what was once a manual and tedious process needing weeks becomes an automated process that takes only a few minutes.
‍

Value in organization

While the organization of images through our data models and pipelines are crucial for streamlining our inspections, the principles of organization extend far beyond just image management. In fact, these principles play a pivotal role in other facets of our work.

In the realm of inspections, especially in hard to access areas, meticulous planning is paramount. One of the challenges we’ve faced is ensuring that our in-field team can safely and efficiently enter these areas. By adopting a data-driven mindset, we’ve transformed this aspect of our inspections. Leveraging previous access notes, publicly available data like topography and road networks, and our proprietary software, we generate optimized flight paths and access maps for our in-field team.

Satellite imagery showing transmission line easement and several possible access routes

As a result, our in-field team is equipped with precise access locations. They also have alternative access points at their fingertips, complete with details about the last usage and the latest conditions. This bounty of information isn’t just for efficiency; it’s about ensuring our team is well-informed, building a repository of access plans for future inspections, and fostering a safer work environment. With backup routes for unforeseen challenges, we reduce field time and enhance safety. As we delve into increasingly complex terrains where detailed planning is crucial, the significance of our data-driven approach becomes even more evident.
‍

What is next

Now that images are organized and managed within the platform, they are enhanced and annotated by our Powerline Technicians. In the next post of the series, we will dive deeper into the process of annotating images and identifying defects and deficiencies.

This post was part of a series detailing Detect’s data journey, produced in partnership with Newfoundland & Labrador Hydro and Stantec.

No items found.

Overall Project Stats

  • Structures Inspected: 1314
  • Photos Taken: 39,401
  • Defects Detected (compared to previous inspection): 10X
  • Total Distance: 220 km
  • Total Inspection Time: 22 Days
  • In-field Team Composition: Pilot, Safety Officer, and Data Manager
    ‍
“Management of many is the same as management of few. It is a matter of organization.” — Sun Tzu


In this third installment, we’ll discuss data uploading and data organization. Given the 1314 structures and +39,000 photos, it is important to build robust and automated systems to handle this volume of data. This project involved two parallel transmission lines, L23(A) and L24(A), spanning approximately 220 km between Wabush and Churchill Falls, NL.
‍

Data Models and Entity Relationships

First and foremost, understanding our data practices is crucial, as they have empowered our team to conduct inspections both safely and efficiently. Data models serve as the foundational blueprints for our entire data management system, streamlining data organization and specifying relationships between distinct data elements. Entity-relationship diagrams (ERDs) further enhance our comprehension by visually mapping out these specific connections.

Imagine you are managing a city’s transportation system. You would need to manage roads, public transit, and bike paths ensuring each neighbourhood is accessible. You would need to make decisions on the best means of transportation across these different routes, understanding the changing needs of your city. Similarly, a data model looks at building a coherent structure, determining how data pieces fit together and relate to one another for efficient access and organization.

ERDs are akin to a detailed city map, showing how these unique modes of transportation come together. A small road may be connected to a major road, acting as the main artery through your city. Bike lanes, bus stops and train stations may intersect or be alongside this major road. These unique elements all interact with each other, with specific rules and frameworks to ensure safe and efficient modes of travel. An ERD looks to do the same thing with data, highlighting specific linkages between different data elements so that we can navigate our data infrastructure seamlessly. A well designed data model and ERD standardize how different data entities relate and interact with each other.

To delve deeper, consider a structure as an example: while multiple images can be associated with a single structure, and several images might reference multiple structural members, a structure is uniquely linked to just one powerline. Conversely, a powerline can encompass numerous structures.

It is important to understand the underlying piece as it serves as the foundation for our image pipelines and how it enables the later stages of the inspection. Data’s journey into our platform is managed using our data models.

Relationship between Structure, Images and Defects

The importance of metadata

Once the in-field team has collected the images, these are uploaded onto our platform. It is here where the data model and our pipelines come into play.

First, it is important to understand what data exists when images are taken. When an image is taken using our drones, the image of the structure is made readily available for our annotators to review. The challenge arises when you have +39000 photos, especially when there’s no clear understanding on their associated structure or viewing direction.

Metadata, in its simplest form, is data about data. Think of it like the label on a folder in a filing cabinet; it gives you essential information about what’s inside without having to open the folder itself. This information accompanies every image we take and becomes the key to addressing the challenge. It provides context and helps us organize our images, making the image metadata an indispensable part of our process.

Sample structure image

When images are taken, they are geotagged with the date and time it was taken. Furthermore, with our drones, the heading and angle of the photo alongside the elevation is available. What this allows us to do, using our proprietary software, is automatically associate these photos to any given structure quickly and accurately.

Basic metadata overview

In traditional drone-based inspection services, it’s not uncommon for teams to receive vast quantities of photos in an unorganized manner, necessitating manual sorting and organization. With our data models and software, what was once a manual and tedious process needing weeks becomes an automated process that takes only a few minutes.
‍

Value in organization

While the organization of images through our data models and pipelines are crucial for streamlining our inspections, the principles of organization extend far beyond just image management. In fact, these principles play a pivotal role in other facets of our work.

In the realm of inspections, especially in hard to access areas, meticulous planning is paramount. One of the challenges we’ve faced is ensuring that our in-field team can safely and efficiently enter these areas. By adopting a data-driven mindset, we’ve transformed this aspect of our inspections. Leveraging previous access notes, publicly available data like topography and road networks, and our proprietary software, we generate optimized flight paths and access maps for our in-field team.

Satellite imagery showing transmission line easement and several possible access routes

As a result, our in-field team is equipped with precise access locations. They also have alternative access points at their fingertips, complete with details about the last usage and the latest conditions. This bounty of information isn’t just for efficiency; it’s about ensuring our team is well-informed, building a repository of access plans for future inspections, and fostering a safer work environment. With backup routes for unforeseen challenges, we reduce field time and enhance safety. As we delve into increasingly complex terrains where detailed planning is crucial, the significance of our data-driven approach becomes even more evident.
‍

What is next

Now that images are organized and managed within the platform, they are enhanced and annotated by our Powerline Technicians. In the next post of the series, we will dive deeper into the process of annotating images and identifying defects and deficiencies.

This post was part of a series detailing Detect’s data journey, produced in partnership with Newfoundland & Labrador Hydro and Stantec.