Point Cloud and Raster Change Detection

Summary

In this assignment, students difference point cloud and raster data sets to compare structure from motion (SfM) and terrestrial laser scanner (TLS) results. They interpret error in these methods used for high-resolution topography collection.

Day 8 Morning - This activity is part of the 2-week remote field course Geoscience Field Issues Using High-Resolution Topography to Understand Earth Surface Processes

In spring 2020, the world was hit by a pandemic that spread globally by March, causing universities and most of the world to move to remote means. Summer field camps, long hailed as a rite of passage in the geosciences, were canceled throughout the US. The community moved quickly, with NAGT developing remote learning tools and arranging for sharing and collaboration between instructors and institutions. As such, UNAVCO (GETSI)and University of Northern Colorado embarked on a data collection campaign for a summer field course entitled "Geoscience Field Issues Using High-Resolution Topography to Understand Earth Surface Processes" – originally slated for in-person teaching. The team collected GPS/GNSS data, drone imagery for use in structure from motion, and terrestrial laser scanning from a site near Greeley, Colorado on the Poudre River.

Share your modifications and improvements to this activity through the Community Contribution Tool »

Context

Audience

This exercise is intended for majors-level geoscience courses that have field or remote (online) field components.

Skills and concepts that students must have mastered

Familiarity with point cloud data, geospatial data, raster data and a background in geomorphology and/or physical geography concepts is helpful.

How the activity is situated in the course

This is situated after the first week of SfM activities and the TLS activity, as outlined in the Sheep Draw Vignette and the course page. As such, students have been introduced to point cloud data and associated post-processing techniques, and raster data. Furthermore, students have used CloudCompare and ArcGIS before this exercise. If this were adapted for use without creating an SfM model or creating one's own raster, CloudCompare would need to be introduced and familiarity with ArcGIS required.

Activity Length

The instructor gives two lectures, one on point cloud differencing (~30 mins) and one on raster differencing (~30 mins). The activity could be broken up in to the corresponding point cloud and raster portions of the exercise. For the Point Cloud Differencing lecture, students could follow along with the instructions and conduct the analysis with the instructor. The students work for ~2 hours on completing the assignment. A check-in or office hours for trouble shooting is appropriate.

Goals

Content/concepts goals for this activity

Difference point clouds and rasters used in error analysis and change detection

Higher order thinking skills goals for this activity

Distinguish different applications and interpretations of data based on the context (e.g., error analysis for concurrent space and time versus geomorphic change detection for concurrent space but different time)

Other skills goals for this activity

Using a CloudCompare Plug in; using the ArcGIS raster calculator

Description and Teaching Materials

This exercise uses the students' previously-created SfM dataset (Day 3 of course) and pre-collected TLS data (Day 7 of course) to conduct differencing of point clouds as well as a rasterized version of the same data. Students use their own point clouds and rasters used earlier in the course, but they are also provided. After the instructor gives the lectures on "Point Cloud Differencing", students conduct a point cloud differencing of the SfM and TLS data for their area of interest (Area 1 or Area 2; see Vignette) using CloudCompare. Since these datasets were collected at the same place on the same day, differences between the datasets should be interpreted as error in one or both of the models. Students are asked to interpret the 3D differences between the datasets. The second lecture, "Raster Differencing" discusses best practices in preparing rasters for differencing. Students then use ArcGIS Raster Calculator tool to subtract one raster from the other (from same area of interest as the point cloud portion). Students interpret the results and compare their differences conducted in 3D (point cloud differencing), and 2.5D (raster differencing). Students interpret their results as an error analysis and discuss which dataset they think is more accurate (and why) and which method provided the most robust error analysis.
Sheep Draw Vignette (Microsoft Word 2007 (.docx) 2MB Dec18 20)
Point Cloud and Raster Change Detection Student Activity (Microsoft Word 2007 (.docx) 1.6MB Dec18 20)

PointCloudDifferencing.pptx (PowerPoint 2007 (.pptx) 15.9MB Jan4 21)

Raster Differencing Lecture (PowerPoint 2007 (.pptx) 5MB Jan4 21)


BETH - need to link to SfM and TLS point clouds from Day 7

Technology Needs

A computer with CloudCompare installed is required.
ArcGIS Raster Calculator

Teaching Notes and Tips

You will need to help students find the appropriate tools in both CloudCompare and ArcGIS. As such, synchronous remote instruction, in-person computer lab teaching, breakout rooms or frequent check-ins would be appropriate


Assessment

As the summative assessment, students answer questions about the differencing methods used. Formative assessment should be done through discussion with students as a whole group or individually.

References and Resources

Adapted from GETSI Unit 4: Geomorphic Change Detection

Explanation of M3C2 cloud differencing plugin

CloudCompare tutorial on cloud differencing

Point Cloud Differencing presentation adapted, with permission, from
J Ramón Arrowsmith (School of Earth and Space Exploration Arizona State University), Ed Nissen (Colorado School of Mines), and Christopher J. Crosby (UNAVCO)

Raster Differencing presentation adapted, with permission, from Philip Bailey (North Arrow Research), James Brasington (University of Waikato) and Joe Wheaton (Utah State University)



Comment? Start the discussion about Point Cloud and Raster Change Detection