Assessing the Quality and Accuracy of Geospatial Data

Assessing the accuracy and quality of geospatial data has always been a focus for ASPRS. With time, the sensors, processing capabilities, and needs of users all have changed, requiring new approaches to measure and ensure the quality and accuracy of geospatial data sets. This session will feature advancements and approaches towards measuring spatial data accuracy for lidar, UAS, satellite, and combined data projects.

Measuring the Consistency of Lidar Data at Large Scale Using Cloud Computing and Open-source Software 

This paper describes the use of opensource software and algorithms to estimate the geometric consistency and quality of large scale lidar data programs and in particular, the USGS  3D Elevation Program. We present work on developing algorithms to exploit 3DEP data available on cloud computing platforms using open-source tools for processing large amounts of data. We present three algorithms for assessing the quality of lidar data at project level and within projects.  First, we present methods to determine gaps between two lidar projects, as well as gaps within a lidar project. The identification of these gaps is important for ensuring comprehensive data availability over the US, as well as serving as important tools for project quality validation. We then present an algorithm to compare portions of lidar data that fall in the overlapping regions of the point cloud between multiple projects, based on multiple parameters (elevation, slope, curvature, and roughness). Iterative plane matching algorithms are then applied to determine the alignment of data in these overlapping regions. The information on alignment is crucial in informing downstream scientific applications (such as hydrology) etc. Results of analysis from state level projects such as Puerto Rico  etc.,  collected at different times (after the year 2015) are present. Finally, we present methods to compare 3DEP data to data of higher accuracy  using point cloud registration techniques such as iterative closest plane.  

Aparajithan Sampath, KBR Contractor to US Geological Survey

Practical Application of Cumulative Error in Multi-Sensor Projects

This presentation will demonstrate the utilization of cumulative error calculations across numerous multi-sensor projects.  A sample project may include the error estimate of a photogrammetric product that is registered with the assistance lidar for example.   In aggregate, and by project types, comparisons will be drawn between traditional and emerging methods of error estimation.  Finally, the impacts of selecting a particular method, or methods, on the project planning and customer expectations will be explored.

Ryan Rittenhouse, GPI Geospatial Inc.

Horizontal Positional Accuracy of PlanetScope High-Resolution Imagery

With unprecedented daily revisit capability, Planet Labs high-spatial resolution, and high-temporal resolution PlanetScope imagery holds the potential for many Earth observation applications. PlanetScope data have not been statistically tested in terms of horizontal positional accuracy. In this presentation, the results of an evaluation of the horizontal positional accuracy of PlanetScope satellite imagery in comparison to National Agriculture Imagery Program (NAIP) satellite imagery will be discussed. Testing was completed according to the ASPRS Positional Accuracy Standards for Digital Geospatial Data.

Lisa Sinclair, University of New Mexico

3D Accuracy Assessment of Airborne Lidar Point Cloud

Accuracy assessments of airborne lidar point clouds typically estimate vertical accuracy by computing RMSEz (RMSE in z coordinate) from ground check points (GCPs). In most cases, it is limited to the z-coordinate (elevation) because the low point density of the airborne lidar point cloud does not have adequate semantic information. As lidar technology advances, lidar point clouds have higher point density and better precision. This advancement calls for improved data quality assessments. We propose two approaches to achieve full 3D accuracy assessment, one using geometric features, the other using amorphous objects. Geometric feature-based GCPs can be derived from any geometry that determines a unique GCP, such as an intersection point from 3-planes, an intersection point from two lines, an intersection point from a line and a plane, etc.  Retaining walls and parapet-like building roofs make good linear features because the point clouds behave like a linear isolated object.  Intensity-based linear features (paint lines on a road or parking lot) can be utilized also.  An amorphous object-based technique is very useful when man-made objects are not readily available. Ground-scanned point clouds can serve as references to estimate the 3D georeferencing error of airborne lidar point clouds. We show several 3D accuracy assessment examples from recent USGS airborne lidar data collections and ground survey campaigns to demonstrate that advanced 3D accuracy assessments are possible.                

Minsu Kim, KBR

Measuring the Consistency of Lidar Data at Large Scale Using Cloud Computing and Open-source Software 

This paper describes the use of opensource software and algorithms to estimate the geometric consistency and quality of large scale lidar data programs and in particular, the USGS  3D Elevation Program. We present work on developing algorithms to exploit 3DEP data available on cloud computing platforms using open-source tools for processing large amounts of data. We present three algorithms for assessing the quality of lidar data at project level and within projects.  First, we present methods to determine gaps between two lidar projects, as well as gaps within a lidar project. The identification of these gaps is important for ensuring comprehensive data availability over the US, as well as serving as important tools for project quality validation. We then present an algorithm to compare portions of lidar data that fall in the overlapping regions of the point cloud between multiple projects, based on multiple parameters (elevation, slope, curvature, and roughness). Iterative plane matching algorithms are then applied to determine the alignment of data in these overlapping regions. The information on alignment is crucial in informing downstream scientific applications (such as hydrology) etc. Results of analysis from state level projects such as Puerto Rico  etc.,  collected at different times (after the year 2015) are present. Finally, we present methods to compare 3DEP data to data of higher accuracy  using point cloud registration techniques such as iterative closest plane.  

Aparajithan Sampath, KBR Contractor to US Geological Survey

Topographic Differencing Between Lidar Projects to Assist in Data Validation Geo Week 2024

The National Geospatial Technical Operations Center (NGTOC) of the U.S. Geological Survey processes lidar data collected for the 3D Elevation Program. This data provides foundational elevation information for earth science studies and mapping applications over the conterminous United States, Alaska, Hawaii, Puerto Rico, and other territorial islands. As we approach 100% coverage of 3DEP quality lidar data, we will have new collections that overlap previous lidar projects. Topographic differencing measures landscape change from a variety of anthropogenic and natural sources such as new building construction, cut and fill operations, erosion, earthquakes, and landslides. With multi-temporal lidar collections, we can perform vertical differencing, which is the subtraction of one grid-based bare-earth digital elevation model (DEM) from another. This can show true change as well as accentuate errors in the input datasets. This presentation will discuss the concept of topographic differencing and show examples of real change due to natural and anthropogenic causes and apparent change that is actually noise caused by georeferencing errors, flight alignment errors, incorrect vertical datums, and inconsistent ground classification.

Barry Miller, United States Geological Survey

Assessing the Accuracy of UAV Aerial Surveys

Recently, developments in airborne sensors and easy to fly, reliable, low-cost commercial UAVs have opened a new era for precise aerial mapping. Photogrammetric mapping principles are employed in collecting, processing, and analyzing optical images collected using UAVs. In this presentation, data collected with different UAV configurations are assessed.  We flew two different sUAVs:  3DR IRIS+ quadcopter and a fixed-wing homemade TuffWing drone. Both drones were equipped with a Canon PowerShot S100 camera and were flown between 20 meters and 127 meters of elevation. GCPs were set with different configurations.  Geospatial products including DEMs and orthophotos produced with commercial software and with different ground control configurations are evaluated. Check-points were then used in the evaluation process. Ground surveying control was collected with a TS02 total station and served as the ground truth for data evaluation. The comparison of the derived 3D information from sensed data with ground measurements showed high correlation between the accuracy of the 3D products and the sensor specification, flying altitude, and image layout.  The accuracy of photogrammetric products changes depending on the number and distribution of ground control points (GCPs), especially with a small number of GCPs.      

Dr. Ahmed Elaksher, New Mexico State University