Pairing semantics and object-based image analysis for national terrain mapping - a first-case scenario of cirques

Abstract

As the new National 3D Elevation Program (3DEP) prepares to provide high-resolution lidar coverage for continental United States, Hawaii, and the territories, it is important to consider that terrain information captured in elevation data are pixel-based instead of feature-based. Referencing 3DEP data for semantic access and inferencing requires the transcription of pixels into accessible features. Indeed, accessing and inferencing terrain features renders them more operational for long-term national mapping. Much progress has been made in transcribing pixels into terrain features using Geographic Object-Based Image Analysis (GEOBIA), as compared to traditional, pixel-based image analysis. However, these studies have focused mainly on European terrain, while their applicability and use for United States transcription has not yet been adequately determined. This research evaluates that applicability relative to the mapping of glacial cirques in the Tahoe Basin using an established GEOBIA workflow, run on eCognition. Results suggest that while some parameters of the workflow may require modification, the general workflow steps may apply to other regions.