GIS/MEA 584: UAS Mapping and Analytics – UAS and Lidar Data: Comparison, Fusion, and Analysis

Outline

Motivation for combining lidar and UAS SfM data
Analysis of lidar and UAS SfM DSMs differences
Patching and smooth fusion of UAS and lidar DSM
Overland flow simulation on fused DEM

Point Clouds from Lidar

Measured variable is the time of return for each pulse
Georeferencing is based on the position (measured by GPS) and exterior orientation (measured by inertial navigation system INS) of the platform
(x, y, z) is derived from time, GPS positioning, and INS parameters

Point cloud visualization from Lidar

Point Clouds from SfM

Measured variable is reflected energy captured as imagery
Georeferencing can be done entirely from GCPs
(x, y, z) is derived from overlapping images and GCPs
Alternatively: GPS and INS for the position and orientation of images and camera parameters

Point cloud visualization from SfM

Comparison of Point Cloud Properties

Airborne Lidar:

Passes through vegetation, multiple returns
Shifts between swaths, corduroy effect
Regional coverage
High cost

UAS SfM:

Very high point density
Limited capability to map under vegetation
Influenced by cast shadows
Low cost but limited area

Different distribution of errors and distortions:
Lidar and SfM provide an independent set of measurements.

Motivation for Lidar and UAS SfM Fusion

Low-cost DEM updates in rapidly changing landscapes (e.g., construction sites)
Replacing vegetated areas in UAS SfM with bare ground
Watershed analysis when only part of the watershed was mapped by UAS
Improve accuracy and level of detail over stable features (buildings, bridges)
Add your suggestion

Fusion Techniques

Merge the point clouds, decimate or bin at high resolution, interpolate: computationally demanding
Interpolate DEMs, randomly sample, patch, and reinterpolate: loss of detail, time-consuming
Interpolate DEMs and patch: sharp edges
Smooth fusion - requires DEMs at the same resolution

Workflow

Evaluate the point clouds, select resolution
Interpolate lidar and UAS DEMs at the same resolution
Evaluate the differences between the lidar and UAS DEMs
If applicable, identify errors and apply corrections
Apply smooth fusion

Identify Distortions Due to Processing

Identify the spatial pattern of distortions
Difference maps between Lidar DSM and UAS SfM DSMs

Agisoft

Pix4D

Identify Changes in Topography

Vegetation growth or removal
Change in elevation surface due to erosion processes
UAS SfM - lidar difference maps with custom color

Comparing Profiles: Fields

Identify systematic errors (vertical shifts)
Differences due to vegetation growth, erosion/deposition

Profiles Comparison Fields

Comparing Profiles: Building

High accuracy, lower level of detail in lidar data
Negligible systematic error

Profiles Comparison Building

Comparing Profiles: Road

Relatively easy to identify stable features
High accuracy - differences between lidar and UAS SfM within 4cm

Profiles Comparison Road

Correct for Systematic Error and Distortions

Systematic error: if due to GPS shift, use median difference
Systematic error: if tilt, use regression function
Complex UAS SfM distortions can be reduced using interpolated GCP differences or differencing with lidar
Importance of well-distributed, accurate GCPs

Updating Lidar DEM by Patching

Replace the grid cells in the updated area by UAS DEM
Average over an overlap
Usually creates an edge - requires post-processing

Images by Brendan Harmon

Updating Lidar DEM by Patching

Edge between lidar and UAS SfM DEMs

Edge Visualization

Smooth Fusion

Derive overlap raster
Weighted smoothing based on the distance from the edge

Fusion Method

Linear combination of elevation surfaces \(z_{A}\) and \(z_{B}\) with weights given by overlap width \(s\) and distance \(d\) to the edge of \(z_{A}\):

\[ z_{AB} = z_{A} w + z_{B}(1 - w), \\[10pt] w = f(s, d) = \begin{cases} \frac{d}{s} & 0 \leq d < s \\ 1 & d \geq s \end{cases} \]

Influence of Overlap Width \(s\)

Overlap Influence

Fusion with Spatially Variable \(s\)

By taking into account spatially variable differences \(\Delta z\) between DEMs \(A\) and \(B\) along the overlap:

We get a more gradual transition where differences are high
We preserve subtle features of both DEMs where differences are small

Fusion with Spatially Variable \(s\)

Absolute elevation difference between lidar and UAS DEM controls the width of the transition overlap.

Fusion Visualization

Lidar and UAS SfM Patching and Fusion

Simply patched vs. fused UAS and lidar DSMs with overlap width \(s = 20\ m\)

Fusing Vegetated Patch with Bare Ground

Vegetated areas in UAS DSM are replaced with lidar DEM. Yellow shows bare ground in UAS DSM.

Spatially Variable Overlap

Spatial Overlap

Bare ground UAS DSM is used to update lidar bare ground DEM using spatially variable overlap

Modeling Overland Flow

Microtopography captured at ultra-high resolution by UAS SfM poses special challenges to flow routing:

Real depressions are important features
Complex pattern of ponding, overflow of barriers is not supported by standard tools
Noisy surface requires a robust algorithm