How to Extract Maximum Value from Lidar Datasets
Government agencies and asset owners around the world are now looking to monitor the natural and built environment at scale, without having to compromise on data quality or schedules. The way we acquire, process, and manage Geo-data provide them with important insights into what the future may hold.
Climate change, population growth and urban expansion all contribute to a range of issues that are exposing people to functional, physical or financial risk. Extreme weather events, floods, subsidence, coastal erosion and deteriorating infrastructure are just some of the growing number of challenges that government agencies and asset owners are having to tackle.
For years they have relied on lidar technology to map the assets and features that fall within their sphere of responsibility –infrastructure including houses, flood plains, trees, rivers and more. Until now, many have relied on a single lidar dataset as their source of truth.
However, the natural and built environment is changing at a rapid rate, so clients need reliable, current data to inform their decisions. A well-classified, high-quality lidar dataset is a rich source of information about the conditions of an area at the time of the survey, but it does not enable users to see what has changed, recognize trends or make predictions. Achieving all of this requires recurring lidar datasets.
Today’s clients want more…
Government agencies and asset owners are now hungry for more data, higher accuracy and more frequent surveys covering more area. They want to be able to compare datasets more often, to identify quickly, easily– and visually – where changes are occurring. They want to be able to model possible future scenarios more precisely, so that they can manage the environment, infrastructure and associated safety risks more effectively. And they want to do it at scale.
But clients don’t usually have exorbitant funds at their disposal. All too often, budgetary constraints thwart their far-reaching ambitions, which then boil down to an unsatisfactory trade-off between the level of financial investment and the quality and accuracy of the data.
Put existing lidar datasets to good use
At Fugro, we acquire and process lidar data throughout the US for five separate programs. There is an abundance of lidar data available. We recognized that by adding or improving classifications to the existing data, users could get more value from it.
To put this theory to test, we selected a canal located near Houston and performed a water impact comparison of two lidar-derived datasets: one current and one acquired four years previously. We created classified point clouds and hydro-flattened digital elevation models from both datasets to characterize features and reveal the shape and contour of the bare earth and how water naturally flows through it.
We then calculated elevation variations by comparing the two datasets. A color-coded raster product highlighted important changes that had occurred during the four years between the lidar surveys. It pinpointed where extreme water flow had eroded the canal banks and where sediment had been deposited, as well as how the environment had altered.
Figure 1. Comparing changes between two lidar datasets. In this example, the vegetation decline is colour-coded in black.
A key element in processing the two datasets was the availability of point classifications for trees and bushes, culverts and buildings (a State of Texas standard deliverable). These features are not standard in publicly available federal programs(USGS 3DEP), and were obtained through re-classifying the data to high accuracy. Visualizing the datasets and changes in 3D helped the team to assimilate and understand the impact of the changes on communities, vegetation, utility and transportation infrastructure.
Figure 2. Simulating flood scenarios with the help of a well-classified lidar pointcloud
Broad application potential
Using traditional methods, lidar point cloud classification can be a laborious, time-consuming and expensive exercise that relies heavily on human expertise to label the data. It is possible to achieve high-quality, accurate results, but only if clients have a generous budget and ample time at their disposal.
It is now possible to transform the lidar point cloud classification process by combining human expertise, cloud-based processing, automation and machine learning. Point cloud data can be classified almost 40percentfaster and at scale, while improving the quality and accuracy of the dataset.
Although used to address issues relating to erosion, sediment deposits and flow in the Texas example, machine learning lidar point classification techniques can also be used to accurately geo-locate and delineate structures, identify areas that are prone to flooding, monitor the density and health of vegetation over time, and examine the impact of climate change on ecosystems.
Although there’s no denying the growing importance of acquiring new and recurrent lidar datasets, there is also a very strong argument for making the most of the data that’s already available. It’s a faster, safer, more sustainable solution that can add significant value.