Work

From photogrammetric system development and data analysis to game engine visualization

Year: 2026
Type: dissertation
Abstract: This article-based doctoral thesis advances an end-to-end workflow that connects photogrammetric system development and data analysis to interactive, geospatially faithful visualization in a modern game engine. The objective is to increase the operational value of image-based 3D reconstructions by making quality-controlled datasets portable across analytical environments and real-time rendering platforms, thereby improving the analysis, communication, and reuse of geospatial products. The thesis builds on scriptable photogrammetric quality control that iteratively prunes sparse reconstructions by removing high-reprojection-error points, typically about 1% per iteration, until residuals fall below half a pixel, yielding a controllable foundation for automation in production contexts. Application studies include a qualitative assessment of urban green infrastructure (UGI) with a custom drone platform, a comparative analysis of airborne laser scanning (ALS), and UAV photogrammetry for tree-level above ground biomass (AGB) estimation in urban settings, establishing modality-specific trade-offs and the value of hybrid data use. The main contribution illustrated in Paper V is a practical pipeline that carries geospatial analyses from desktop tools into Unreal Engine for real-time, geographically anchored exploration. Specifically, outputs such as a watershed analysis from QGIS and change detection from CloudCompare are integrated and visualized interactively, including lighting studies using Unreal Engine’s Lumen technology to examine illumination scenarios without sacrificing spatial context. This workflow demonstrates how analytical results can be rendered as an interactive “digital twin”, while noting that maintaining digital twins requires regular updates that were conceptually addressed but not implemented within the scope of the thesis. The work also clarifies when point clouds should be favored for preserving radiometry and fine structural detail and when mesh-based representations are preferable for performance and surface continuity in visualization tasks. Across the studies, the thesis underscores three conclusions. Firstly, robust, scriptable quality control in photogrammetric reconstruction enables repeatable, production-grade pipelines that preserve traceability from acquisition to visualization. Secondly, for urban forestry applications, UAV photogrammetry contributes meaningfully to AGB estimation but exhibits systematic differences relative to field measurements and ALS, reinforcing the concept that no single modality is universally superior and that hybrid approaches are recommended. Acquisition conditions (e.g., leaf-off, or wind) can differentially affect modalities and analysis tasks such as individual tree detection. Thirdly, bridging analysis and communication through game engine visualization expands the utility and audience of geospatial products, supports circular data practices by reusing analytic outputs in new contexts, and provides a pathway to operational digital twins coupled with update mechanisms. This thesis thus demonstrates a methodology rooted in controllable photogrammetric quality assurance and extended through interoperable geospatial tooling to deliver interactive, context-rich visualization of analytic outputs. By integrating QGIS- and point cloud-based analyses into Unreal Engine and leveraging real-time lighting via Lumen, the work highlights how geospatial accuracy, visual fidelity, and interactivity can be combined to support decision-making and stakeholder engagement in urban environmental applications.
Institution: Aalto University
Language: English
Subfield: Geology
Open Access status: green
OpenAlex ID: W7172474993
Cites: 104
Cited by:
Related to: 10