From comparison to combination: Street view imagery (SVI) and 3D model-based analyses for urban visual environment assessment

Street view imagery (SVI) is widely used in urban visual analysis and often treated as equivalent to eye-level perception. Yet its limitations and contextual applicability remain underexplored. This paper conducts a diagnostic viewpoint-level comparison of an image-based SVI pipeline and a 3D model-based field-of-view (FOV) method to clarify their respective weaknesses, strengths, and how they can be combined in practice (rather than treated as interchangeable or numerically fused). Using the West Lake ring road in Hangzhou as a case study, we analyze 2140 panoramas at 1075 viewpoints. The comparison shows systematic differences: SVI produces higher green shares (+0.16 on average), while FOV yields higher paved ground (+0.13) and building shares (+0.08). Sky differs little overall, water remains minor, and cross-method consistency varies by segment; SVI displays greater local variability linked to canopy occlusion and near-field heterogeneity. A small perception survey validates these findings. Terrain relief and building height were recognized more consistently in FOV, while vegetation and water abundance aligned more closely with SVI. Participants also judged overall ambience more easily from FOV’s structural stability, even though SVI conveyed greater visual realism. These results reveal clear complementarities: FOV provides structure-aware metrics, SVI emphasizes appearance cues, and neither alone captures lived perception. On this basis, we propose a combination-oriented three-layer workflow, with perception as a required validation layer to support reliable applications in skyline and openness control, interface and character management, greenery maintenance, and equity assessment.