<![CDATA[This study evaluates the rendering performance of smartphone LiDAR-derived 3D models in a web-based CesiumJS environment using a mural dataset. Three formats were compared: raw OBJ, monolithic glTF, and 3D Tiles. The geometry comprises 801,949 vertices and 1,372,979 triangles. Asset-level measurements show that transferable payload decreases from 77.82 MB (OBJ) to 67.15 MB (glTF) and 13.67 MB (3D Tiles), demonstrating that tiled delivery offers superior storage and transfer efficiency. The 3D Tiles package includes 89 GLB tiles with a dominant structural concentration at traversal depth 3. Tile-level statistics reveal payload variability between 5.54 KB and 481.88 KB. To evaluate runtime performance, the study proposes a regression-oriented benchmark framework using loading time, frame rate, tile-load intensity, browser type, and device class. Analysis suggests that loaded-tile intensity influences frame-rate variation more significantly than loading time. Findings indicate that 3D Tiles is the most suitable representation for CesiumJS when prioritizing progressive refinement and scalable access. Furthermore, the results confirm that low-cost smartphone LiDAR combined with open-source web technologies is effective for the online visualization and dissemination of geospatial digital twins.]]>
Copyrights © 2026
