高密度商业街区街道空间失序研究与优化OACHSSCD

With the continuous deepening of urban built stock space renewal,street environments face issues such as unbalanced interface organization and visual information interference,which lead to spatial disorder and impair street spatial quality.Compared with general streets,high-density commercial districts are characterized by commercial frontages,dense shop signs and advertisements,and the frequent presence of temporary spatial elements.Multiple street elements are often highly superimposed within limited visual fields.As a result,physical disorder in such areas is no longer merely reflected in direct disorder cues such as graffiti,damage,or waste but is more closely associated with the imbalanced combination of multiple spatial elements and the amplification of their differences across viewing directions.Existing studies have made progress in physical disorder identification,street-view image measurement,and spatial distribution analysis.However,the relationship between the composition of spatial elements,their directional differences,and physical disorder evaluation in high-density commercial districts remains insufficiently discussed.In this study,a research framework composed of physical disorder evaluation,spatial element identification,and relationship analysis was established based on a case study in Sanlitun,Beijing to examine the characteristics and optimization of street spatial disorder in high-density commercial districts. An area covering 1.5 km × 1.5 km in Sanlitun was used as the empirical case.Sampling points were placed at 100 m intervals along the road network based on OpenStreetMap road network data,thus producing 326 sampling nodes and 939 multi-directional street-view images.First,a spatial disorder evaluation system was constructed using MIT Place Pulse street perception data,Q-score-based candidate sample screening,manual verification and labeling,and CLIP-based image-text encoding.Then,a semantic segmentation model trained on the ADE20K dataset was applied to identify pixel-level elements in street-view images.Based on previous studies on spatial disorder,the semantic labels were classified into five categories:natural and visual buffering elements,built spatial background elements,traffic and dynamic interference elements,infrastructure and vertical interface elements,and visual clutter and non-structural elements.At the node scale,the average proportions and directional differences of the five element categories along the front,back,left,and right street-view directions were calculated.Moreover,the correspondence between spatial disorder evaluation and the composition of street spatial elements was disclosed through spatial mapping,comparison between high-and low-disorder groups,and locally weighted regression. The results show that the AUC of the spatial disorder evaluation model was 0.96 on the ROC curve,indicating strong discriminative ability.The spatial disorder scores in the Sanlitun area present a spatial pattern of locally clustered high values and continuously distributed low-to-medium values.High-disorder road segments are mainly located along peripheral arterial roads,district transition areas,and some internal connecting streets,while low-disorder ones are more commonly found in internal secondary streets.Built elements remain at a high level in both high-and low-disorder groups.Natural elements account for a higher proportion and show a more stable distribution in the low-disorder group,indicating a certain environmental buffering effect.Visual clutter elements are more prominent in the high-disorder group and show a more direct correspondence with higher physical disorder evaluation.Directional differences among the four views of the same node are more sensitive to spatial disorder evaluation.The study indicates that spatial disorder in high-density commercial districts is not the result of a single object or a single element but the collaborative outcome of multiple spatial elements,amplified directional differences,and imbalanced interface organization.Based on these findings,peripheral arterial roads should focus on mitigating rigid entrance interfaces,excessive facility overlap,and boundary-related spatial pressure.Internal framework roads should strengthen the integration of non-structural visual elements,such as shop signs,wayfinding systems,outdoor displays,fences,poles,and utility boxes.Local nodes and living streets may improve environmental transitions through corner organization,small-scale greenery supplementation,and pedestrian space repair.This study provides a methodological reference for identifying problematic street segments,optimizing street interfaces,and regulating hierarchical designs in high-density commercial districts.