Web-based Topographic Differencing of High Resolution Topography Data

High resolution topography is a powerful observational tool for studying the Earth's surface, vegetation, and urban landscape, with broad scientific, engineering, and education-based applications. The sub-meter resolution imagery is collected with laser and photogrammetric techniques on ground, air, and space-based platforms. When areas are imaged two or more times, we can perform multi-temporal topographic differencing to detect surface change from a variety of tectonic and geomorphic processes including earthquakes, volcanic eruptions, flooding events, and landslides.

OpenTopography is an NSF-funded facility providing open access to point cloud and raster topographic data and processing tools to generate and visualize derivative products. We are adding the capability to perform on demand topographic differencing for overlapping high resolution datasets in the OpenTopography archive. In this processing workflow, users have the ability to select two overlapping datasets for differencing and will be presented with suggested processing parameters to optimize the differencing results. Advanced settings will allow users to alter the processing parameters. The differencing results are displayed to the user and can be downloaded along with metadata reflecting both datasets and the differencing processing workflow.

We are implementing two types of topographic differencing in OpenTopography. (1) Vertical differencing is the subtraction of raster-based imagery. It can be performed on original raster imagery or on gridded point cloud imagery, but both datasets must be on identical grids. Error analysis is possible by manual assessment of differences in unchanged areas. (2) In phase two of this work, 3D differencing is performed with a windowed implementation of the Iterative Closest Point (ICP) algorithm. This calculates the best rigid deformation (translation and rotation) to align windows of point cloud topography. The window size sets the resolution of the surface displacements and is trade-off between a large scale with enough topographic relief to produce an accurate alignment and a small scale with a lower violation of the rigid body assumption. We optimize the window dimension based on the spatial resolution of the imagery and the topographic relief of the individual datasets.