Rolling Shutter: The Drone Mapping Problem No One’s Talking About

Pix4D
The Science of Drone Mapping
4 min readApr 21, 2016

--

The idea of creating maps and models from drones is no longer a novelty: businesses are now making money from drone mapping services, with more and more people exploring how to do so. But with what drones? The line between “professional” and “prosumer” drone labels is increasingly blurred as stiff competition between manufacturers pushes hardware development. Photogrammetry software walks that blurry line in an effort to produce professional results for both.

When choosing a drone, things like budget, flight time and software were probably on the forefront of your mind: not whether the drone’s camera had a rolling shutter or global shutter.

And yet, using a rolling shutter camera can affect your mapping accuracy

GoPro, DJI Phantom and Inspire1 cameras, among other models on popular drones, all have a rolling shutter. In fact, over half of Pix4D users are processing images from a rolling shutter camera.

What is a Rolling Shutter?

Global and rolling shutters have different ways of exposing the camera sensor to light in order to register an image. A global shutter exposes all of the pixels of the sensor at one time, whereas a rolling shutter exposes each pixel row of the sensor at slightly different points in time.

Think of a rolling shutter camera as a digital scanner or copy machine. You place a piece of paper on the glass window of the machine, where it’s scanned by a moving sensor to create an image copy. The time it takes for the machine to scan one image can be compared to the amount of time a rolling shutter camera takes to register an image: something called read-out time.

Rolling Shutter Distortion

Scanning a piece of paper is a relatively foolproof process as long as you don’t move the paper while it’s being scanned. If you do, the image copy will be distorted in a way very similar to the distortion of an image taken by a rolling shutter camera when mounted to a flying drone.

The Mapping Problem

Imagine you want to map a building with your drone. You use a flight planning app to create a grid-patterned mapping flight, and the drone takes off, capturing overlapping pictures on each line of the flight grid until the area is visually captured. These images are then processed in photogrammetry software, which extracts 3D geometry from them by matching similar features (keypoints) between images: a technique called structure-from-motion. Getting accurate mapping results depends on a correct estimation of the camera’s position and orientation in flight (its exterior orientation) as well as modeling of the camera itself (its interior orientation).

Images taken with a rolling shutter camera become problematic for mapping when captured from a moving drone. With a rolling shutter camera flying at several meters per second, the distance between the camera’s position when reading the first pixel row of the sensor and the last row can be up to one meter. The resulting image has a distortion that affects the software’s ability to accurately estimate a camera’s exterior and interior orientation, and therefore its extraction of accurate geometry.

Distance between the first and last pixel line can be up to 1m for popular rolling shutter cameras at average mapping flight speed

The Solution

The rolling shutter effect can be corrected for in mapping software by utilizing a strong knowledge of how cameras, drone mapping, and photogrammetry interact. Pix4Dmapper 2.1 is the first commercial photogrammetry software to implement such a correction, with an algorithm that treats each pixel line from a rolling shutter image as a separate image with its own camera orientation relative to drone speed.

This correction makes the estimation of camera positions speed-independent, and as a result, drones can fly faster without losing accuracy.

Even if not strictly related to mapping, it’s also worth mentioning that a drone’s flight speed can be measured purely based on the correct modeling of the image distortion!

Rolling Shutter Pros and Cons

  • The Good: Rolling shutter cameras tend to be less expensive than global shutter cameras, meaning they’re perfect to try out the mapping workflow or if you don’t have the budget to invest in more professional equipment.
  • The Bad: With a rolling shutter camera, your drone mapping results will be less accurate than with a professional-grade, global shutter camera.

What it Means

As the market develops and hardware improves, the rolling shutter distortion could change or even be eliminated. Until then, mapping software will need to compensate for its effect, bridging the gap between “prosumer” and “professional” mapping hardware for the best possible results.

In the end, “the fundamental principle of mapping accuracy” is what really matters: meaning the best accuracy is the one you need for your project. Knowing what that is, along with your long term mapping needs, is the best way to evaluate what type of camera, drone, and software to choose.

Read more about rolling shutter on Pix4D’s website, HERE.

--

--