Waveform datapackets: make the most of your LiDAR data
Thomas Wilson
RPS, Mapping
.
Executive Summary
LiDAR instruments with full waveform digitisers are now the normal for aerial LiDAR Scanners. Custodians of LiDAR datasets are increasingly interested in acquiring waveforms to ensure that they do not miss out on opportunities to get more out of their datasets. In this presentation we explore what waveform data is, how it is stored and related to point clouds, and most importantly, some of the insights aboutwhat we can interpret from the waveform data.
Traditionally, the signal received by the LiDAR instrument is processed and reduced to a cloud of 3D points representing the location of the peak of strong return signals. These point measurements indicate the presence of a physical object that the laser interacted with and which reflected a portion of the laser energy back to the instrument. The intensity of this return signal is also measured. Waveforms are measurements of the return intensity at regular intervals for the duration of the laser pulse and therefore include information about what was happening to the laser in the moments before, and after, the peak of reflection. This can give us some information about what shape the object is or what other objects are immediately adjacent to our measurement point.
Several possible uses of waveform data were explored including individual point waveform data measurements and clusters of point waveform data differences. Different measurements were made of peaks in the waveform such as the width, skewness and kurtosis. Without the waveform data we only have the location and intensity or amplitude of each peak. These extra measurements were examined against the information we already had from the LiDAR. For example we had automatically classified the point cloud into ground, buildings, and vegetation. The vegetation class commonly encompasses other non-ground features such as powerlines, fences and Hills Hoists. In order to remove the non-vegetated features out of the vegetation class we examined the new measurements to see how the vegetation classification could be improved.
When examining the waveform data for a group of ground points we found that the waveform pattern was different for some areas. This indicated that there was a change in land cover between the two regions. One area was bare soil at the edge of a dam and the other area dense, low grass. These simple investigations show that simple metrics from waveform data can be used to segregate the landscape and improve our understanding of the potential sources of error in LiDAR datasets.