Mapping Impervious Surfaces from LiDAR

A 2.101Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE

Einstein once said, "To raise new questions, new possibilities, to regard old problems from a new angle, requires creative imagination and marks real advance in science." The geospatial profession is constantly evolving, demanding innovation and pushing for creative, forward thought processes that continuously support business goals and the widespread application of geospatial technology. By using technological advances we are innovating and creating more cost-effective and efficient solutions than we ever thought possible. Whether you are a subject matter expert or new to the game, it is imperative that you possess an open mind and encompass the willingness to explore all the `what if’ and `why’ factors.

Throughout the last decade, traditional methods of impervious surface mapping have been accomplished mostly with hyperspectral and multispectral satellite data or through aerial photography in which the near-infrared band is often significant for the algorithmic feature extraction process. However, over the past few years, the private remote sensing industry has seen a significant increase in spending on LiDAR and a decreased spending on multi-spectral imagery. With programs like the USGS 3DEP pushing to the forefront, it is now necessary to look at LiDAR as the primary data source for creating foundational data layers that are often needed by local, state, federal, tribal and private sector clients. This market shift provides an initial base line that creates a push for innovation in how all data users identify future scopes of work that includes derivatives from existing LiDAR.

We first approached the need to extract building features from high density LiDAR and spent several months focusing on that task alone. Once we perfected that process, we quickly realized we could use similar techniques for the algorithmic extraction of other features. The research led us down many paths, resulting in a capability to extract impervious features directly from the LiDAR point cloud and intensity data. After doing so, Atlantic finalized the innovative workflow to include the following main phases: mission planning, data acquisition, data calibration, generation of suitable LiDAR derivatives/intensity imagery for impervious feature collection, automated feature collection using propriety software along with scripts, QA/QC, and manual data compilation.

Atlantic had the opportunity to perform this innovative workflow for a project with Jefferson County, Alabama. Within this county lies the largest city in the State of Alabama, Birmingham with a population of 1,128,047. Atlantic mapped approximately 1,123 square miles. Hugging the base of the mountains, Birmingham’s terrain ranges drastically in elevation. This vast elevation change next to an extreme urban area poses a significant challenge for any algorithm designed for feature classification or feature extraction from LiDAR.

Atlantic’s methodology for extracting impervious surfaces from LiDAR works with both newly collected and existing data. To determine the acceptability of LiDAR data for this process, it is important to first understand the mission parameters used to collect and develop the data. The variables of flight altitude, swath width, and flight line overlap are all important. However, the most important variable is point cloud density. Our testing has concluded that this process does not work well with LiDAR data that has a density less than 2 points per square meter. The best results come from LiDAR that is 4 points per square meter or greater.

When considering the quality of the input LiDAR data, it is also important to understand how well the LiDAR data were calibrated. LiDAR calibration is performed on the raw LiDAR data to remove roll, pitch, heading, and other related errors from the raw data. The calibrated LiDAR data is validated using ground control points, to ensure it meets the required horizontal and vertical accuracy standards.

Our process of extracting impervious features from LiDAR begins only after the LiDAR calibration has been completed and verified. The LiDAR classification process plays a key role in mapping the impervious features. Atlantic developed new macros to automate the classification process using GeoCue & Terra Scan software. The design of the macros was biased to enhance the impervious features in the Birmingham, AL area. The algorithmic point cloud classification is followed by a rigorous quality control process and manual classification efforts to further clean up the data.

The next phase of Atlantic’s methodology involves using the classified LiDAR data to develop a series of intensity based orthophotos that are derived from GeoCue. These are not ordinary intensity images that can be created from classified or unclassified data. Instead, these images are created only from the fully classified LiDAR data. The choice of using a classified dataset offers more unique capabilities in developing imagery that are biased to the unique impervious surfaces we intend to extract.

The intensity imagery developed, as well as the classified LiDAR, are ingested into our suite of in-house software (including LP360, ArcGIS– Feature Analyst tool, and other Atlantic-developed tools) to automatically extract the impervious polygons. The biased intensity images are used to perform a supervised classification for each different impervious layer. The resultant data layers are merged and reviewed by a geospatial analyst. Once the reviews are complete, the data are edited to correct any misclassifications and to clean-up any irregular geometric shapes. Atlantic’s routines resulted in us needing to edit less than 3% of the final dataset delivered to Jefferson County.

Atlantic was able to solve an age-old geospatial data need using a new angle–specifically, billions of angles. Utilizing LiDAR as the foundation for creating impervious surface features has helped Jefferson County, AL maximize its LiDAR investment ROI.

"I continue to be amazed by the innovative ideas from Atlantic’s technical team" said Brian Mayfield, Atlantic President and COO. "The ability to add value to a foundational data layer like LiDAR is huge for the geospatial industry. Robert’s solution for developing impervious features from LiDAR will help communities across the United States get maximum value from the upcoming USGS 3DEP and other local / regional LiDAR programs."

Robert Yao-Kumah, GISP is a Senior Geospatial Scientist at Atlantic, leading research and development specifically in geodetic surveying, LiDAR and GIS.
Kimberley Denney is Associate Vice President at Atlantic. She works with Atlantic’s clients to engineer custom solutions aimed at meeting their geospatial and business goals.

A 2.101Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE