Designing Successful Photogrammetry Programs for Enterprise Drone Teams – Skyward

A successful photogrammetry program is more than just the software you use to process the imagery captured with your drone. In fact, that’s the last link in the chain. When Skyward’s Professional Services consultants are asked to help a customer design, build, and train for a photogrammetry program, we start at the beginning — with a blank slate — and walk the client through a process that will help ensure success based on their business needs first. 

This process starts with gaining a thorough understanding of the client’s business objectives by asking questions like:

  • What actionable insights do you need to get from your maps and models? 
  • Do you need measurements, the ability to inspect for anomalies, or both? 
  • Do you need 2D outputs, 3D outputs, or both? 
  • How fast do you need the models processed? Do you have the necessary computing power to process datasets in that timeframe? If not, can your network support transferring images to a suitable computer for processing? 
  • How often will you generate models, and how accurate do they need to be? 
  • Do you need a cloud-based collaborative environment with mark-up and sharing capabilities? 

There are lots of business-related questions to be answered before we even begin to understand the realm of possible technological solutions. Only then do we discuss the sensors, drones, control app, and image-processing software. Let’s take a closer look at the first of these foundational aspects, which is often overlooked.

How to Start an Enterprise Photogrammetry Program

The first step in the process is helping the customer to clearly define their business goals for a drone program. Whether it’s monitoring and measuring stockpiles, creating digital twins of construction projects, building information modeling (BIM), terrain surveys, infrastructure inspection, or any of myriad other photogrammetry use cases, clearly defining what is to be accomplished — and how the success or failure of these efforts will be measured over time — is the first step in any sustainable drone program. Skyward’s Professional Services consultants spend as much time as necessary helping clients paint a clear picture of their goals and ROI metrics before selecting hardware.

This often entails providing clients with an overview of what photogrammetry is, and what it can — and can’t — accomplish. Photogrammetry is the process of collecting geo-referenced, two-dimensional images and using software to stitch them together into two-dimensional (2D) maps (orthomosaics) or three-dimensional (3D) models. 

A processing engine like Pix4D uses location data associated with each image to position the images properly and identify the overlap between images. Using this data, the engine estimates the third dimension of topographical areas and vertical structures. The outputs are a 3D point cloud and a photorealistic mesh that serve as a digital representation of the area or structure being surveyed. Teams can use this to gather measurements, inspect surfaces for defects, and gain other valuable insights.

Today, most modern drones are equipped with high-quality cameras and use GPS sensors to tag the images with fairly accurate (meter level) location data, but these may require one of several geo-rectification techniques using higher end geo-tagging to achieve centimeter-level accuracy. These capabilities, paired with their aerial perspective, make drones ideal for photogrammetry surveys, and enterprise drone programs can recognize the benefits of photogrammetry right away. 

In many industries, aerial mapping and 3D modeling can be part of the value proposition when calculating the return on investment (ROI) in a drone program, but successful photogrammetry requires safe operation of the drone and effective collection of the imagery. And learning to do that while avoiding a significant amount of trial and error requires photogrammetry training, which is what Skyward Professional Services provides.

A common mistake when standing up a new program is to start by selecting the drone, but that selection is actually much further down the decision chain than most initially assume. Why? Because the top priority is to have a clear understanding of the data requirements necessary to satisfy business needs, and work backwards from there. 

Only after the client has a solid grasp on the insights they need — including factors like the accuracy of measurements, frequency of modeling, and how the raw data and processed models will be managed, shared, and analyzed within the organization — can they move on to looking at what hardware will meet those needs. But that still doesn’t necessarily mean a drone. No, understanding these variables only helps one determine the specifications of the sensor that will be used to gather the needed imagery.

Evaluating Sensors for Photogrammetry

Determining the sensor that will meet your photogrammetry requirements begins with defining how accurate you need your images and measurements to be in order to meet your previously determined business goals. 

We define accuracy by comparing the camera’s image resolution with the stand-off distance between the sensor and the target, and by the geospatial accuracy of the GPS tagging of each image. Geospatial accuracy is determined by the quality of the GPS sensor used to tag each image, which will affect the accuracy of measurements derived from the model. 

Cameras with high-resolution sensors will produce clearer, less grainy models, allowing for better and more accurate visual inspection and anomaly detection. However, the further a drone flies its camera from the target being modeled (e.g., for obstacle avoidance or to capture a wider area) the higher the resolution is needed to maintain an adequate spatial resolution. 

A camera’s spatial resolution is also referred to as its Ground Sample Distance, or GSD, and is commonly measured in centimeters per pixel (cm/pixel). GSD is defined as the area on the ground that is represented by each pixel in an image, and it is a function of camera resolution and stand-off distance. For example, an image taken by a 12 megapixel (12MP) camera has 3,000 x 4,000 pixels. Those pixels will be distributed evenly over any given area in the camera’s view-finder such that a drone with a given camera pointed down (also known as a nadir camera angle) may generate a crisp image at tens of feet above the ground, but a grainy and blurry image at hundreds of feet above the target.

Putting it All Together

Once the client’s accuracy requirements are defined, a camera can be selected. This simultaneously involves selecting an appropriate drone to carry it. In many cases, a single off-the-shelf solution readily presents itself. (For example, we’re excited to see the capabilities of the built-for-photogrammetry, 4G LTE-connected Parrot ANAFI Ai.) The same can be said of the image-processing engine. 

But often our Professional Services consultants can accelerate these decisions by sharing their subject-matter expertise and leveraging our massive fleet of aircraft to help the client test and assess different platforms to find the one that most precisely fits their needs. This can include test flying a variety of drones and payloads over representative target types and using different autoflight controllers to gather the datasets needed to meet the business case and generate ROI. From there, we help organizations evaluate data-processing software to find the one that will be the best fit for their application. 

All of this comes back to Skyward Professional Services’ initial goal: helping customers understand and define their business needs — what problems are they trying to solve, and what data is required to solve them — and working from there to get them the hardware, software, and training necessary to meet that business need.

Skyward Guide: 4 Keys to Good Maps & Models