Integrated Analysis of Visual Change Points along Pathways: Automation and Comparison with Image Segmentation and Isovist Representation

This study presents a methodology integrating image segmentation, optic flow, isovist analysis, and change-point detection to identify significant visual changes along a predefined route automatically. A spiral-like path covering a one square kilometer area was designed to detect inconsistencies in routes regardless of linear distances between start and end points. By generating points along this path using various modalities, including random distribution and fixed step, this study aimed to identify inconsistencies independent of point density. The isovists were generated from each point of view, using obstacles such as buildings, walls, and trees, employed for spatial analysis. Integrating street view images and image segmentation techniques enabled comprehensive visual feature analysis. Optic flow analysis further enriched our understanding of dynamic visual changes. Identified change points were validated using dynamic time warping (DTW), revealing their deviation from the path's trend. This study demonstrates the effectiveness of spatial analysis, isovist analysis, and computer vision methods in understanding visual dynamics in urban environments.