The perception of beauty, though often subjective, is influenced by identifiable structural and spatial patterns that shape how individuals experience their surroundings. This study explores the roles of object variety and connections in scenic images in shaping perceptions of environmental aesthetics, using advanced computer vision techniques and regression analysis. Drawing on data from the Scenic-Or-Not project and leveraging the Segment Anything Model, we analysed landscape photographs to understand how object diversity and spatial arrangement affect aesthetic judgments. Our findings reveal a positive correlation between object diversity and perceived scenicness, emphasizing the importance of visual richness and complexity in enhancing scenic appeal. However, excessive object diversity can introduce visual clutter and diminish aesthetic value. Our analysis of object connections, measured through graph-based metrics like network density and clustering coefficient, reveals that denser and more interconnected arrangements enhance scenic appeal, while overly efficient local connections reduce visual interest. These results demonstrate the importance of balancing complexity, coherence and interconnectedness in scenic design. By situating these findings within established theoretical frameworks, this study provides insights for disciplines such as environmental science, urban planning and landscape management, offering guidance for creating environments that evoke positive aesthetic experiences while maintaining visual harmony and interest.
Keywords:
artificial intelligence; environmental aesthetics; image segmentation; scenic perception; spatial networks.
