Not all LiDAR is equal – Part 3: Swathes and how to get them working for you
This week Anditi’s principal software architect and LiDAR dataset diagnostics guru Patrick Poissant and Head of Business Operations and quality queen Catherine Pepper continue the 4 part blog series on the common pitfalls to avoid in the capture of LiDAR with a discussion on things to consider when planning your capture program. See ‘Part 1: Planning for success’ for ways to plan your project to ensure you’re setting yourself up to succeed, and ‘Part 2: The Role of the IMU’ for what factors you should consider to ensure the data reporting from the IMU is as accurate as possible.
This post focuses on another key element of data collection – swathes and how to get them working for you, including best practices for flight paths, swathe overlap and point density for better results in the processing of your LiDAR data.
What is a swathe?
A swathe is the spatial arrangement of returns generated by a pulse from a remote sensing scanner. Often also referred to as the ‘flight line’ or ‘strip’, the swathe is the set of points collected by a single scanner, overlapped with other swathes to generate a dataset.
LiDAR collection has traditionally been thought of as costly, and therefore capturing as little data as possible has been one of the methods used to reduce flight time, and in turn, decrease cost. However, our experience has shown that capturing the bare minimum often leads to holes in the data, which creates a dataset that was cheap to capture but is difficult to classify, and of not much use for 3D analytics and providing accurate spatial insights.
The best way to manage cost and quality of the data captured is to effectively plan your flight path. This involves a few steps that we’ve outlined below.
1. Plan a flight path to ensure you will get 100% coverage of the area you’re interested in
First things first – identify the area you’re interested in! It’s important to allow a buffer for coverage of the exact area you want, as wind and weather conditions can often put your flight plan off slightly. Covering a little extra gives you margin for error that won’t result in a compromise on a complete data set. We recommend a buffer of a few hundred metres around your area of interest.
2. Allow for overlap
Our experience shows that flight paths that allow for approximately 30-40% overlap of the swathes are likely to give you the best results as that helps prevent gaps in the data. Using this amount of overlap gives the data processing platform you’re using the best chance of recognizing patterns in the data, and using them to align the swathes. However, increased point density does mean increased processing time and demand on the processing system, something Anditi has overcome with patented technology and smart distributed computing. In fact, in some cases an increased point density may result in reduced data processing costs because features will become easier to detect and therefore require less pre-processing and computation.
3. Plan for the appropriate point density
As we described in the first post in our series on LiDAR point clouds, planning your project based on what you want to do with the data is key. This is particularly the case when deciding on the appropriate point density for your project. The point density you need is determined by the size of the near ground features that you are trying to detect. The spacing of LiDAR points needs to be at most half the size of the feature you want to detect. For example, in the case of malleefowl mound detection, the ground features we are trying to detect have a radius of approximately 1 to 3 metres. That means you need a point density of at least 2 points per square metre to be able to detect the shape of the mound with any confidence. Confidence in detection of mounds increases with increasing point density, so we generally recommend point densities in the order of 6 to 10 points per m2. You can read about our malleefowl mound detection work in the Murray Sunset National Park Case Study.
The point density is a function of many different decisions in the LiDAR collection process including:
- Aircraft Elevation
- Scanner Frequency
- Scan pattern
- Overlap strategy
Flying at a higher elevation results in cheaper acquisition of data because the width of the swathe will be larger and the area can be covered more quickly. However, most scanners experience a slight reduction in precision associated with the greater distances between the scanner and the ground. Optimising the combination of these factors requires careful planning and should include discussions between the data collection team and the team who will be doing the analytics to ensure the data captured is fit for its intended use.
4. Post processing is key
Technology has come a long way, and understanding the scanning frequency, or point density, you need for your job is important. Traditionally satellites have collected data at 50cm resolution or more. New technology means data can be collected by airborne lasers in excess of 100 points per m2. Of course such a high density of points can mean extremely accurate DTMs, 3D models and analytics, but it also means huge datasets that require smart handling so as not to overload your system. Again this is where we pride ourselves on having some of the best high performance computing technology on the market.
This is why it is important to know what you need for your project, the less density you need in data capture, the less resolution you will have, but the cheaper it will be. 30cm pixels are fine for identifying broad features such as new developments and bodies of water, but higher densities are helpful when you’re using automated classification to identify features like roof planes and powerlines.
The beauty of well aligned swathes
If you manage to properly plan your flight path, create a nice overlap in your swathes and have access to a nice spatial data processing platform, it is possible to align the data from each swathe so accurately that the resulting union acts as a single, accurate point cloud, with double the point density of each individual swathe.
What to do if your swathes are misaligned
Over the years we’ve seen examples of datasets where swathes are misaligned, swathe overlap is insufficient, or the dataset is of a quality that makes it difficult to achieve the kind of outcomes that were hoped for. If you’re in that boat, we have developed a series of tools and processes to review the data, identify any problems and determine if and how the dataset can be fixed. Send us a message via our Contact Us page or give us a call on 1300 326 170 to find out more about how our data diagnostic tools and value added services can help.