Abstract: In the context of national support for rural revitalization and the development of high-quality characteristic towns, this people-oriented microscale study involved real-time observation and objective quantization of path space quality, physical environment measures, and tourist behaviors in a tourism-dominated characteristic ancient town. The specific influencing paths underlying the relationships between path space quality, physical environment measures, and tourist behaviors were analyzed. Moreover, the mechanisms underlying the influences of the multi-dimensional space properties and physical environment on tourism behaviors were discussed. The results not only offer several suggestions and scientific references for the optimized configuration of tourist path spaces and the improvement of space vitality in ancient towns, but also facilitate the development of the cultural tourism industry in towns with Chinese characteristics.
　　A case study of Sanhe Ancient Town in Hefei City was carried out. Two-dimensional and three-dimensional field models of Sanhe Ancient Town were built using unmanned aerial vehicles (UAVs). Using the ArcGIS platform, the field models were overlapped and aligned with an open street map. Sampling points were set through GIS and the geographical coordinates of the sampling points were acquired. The geographical coordinates of the sampling points acquired by GIS were matched with the field models to obtain comprehensive information, including the locations and surrounding environments of the sampling points in real life. Moreover, a 30-m diameter unit research scope centered at each sampling point was established. After the locations of the sampling points were determined, panoramic cameras were fixed at these points to capture street panoramas which were then input into a convolutional neural network model and Open CV algorithm through projection transformation and standardization to obtain the proportion of spatial elements, space colors, and other information within the research scope. Point of interest (POI) information for Sanhe Ancient Town was collected by accessing Amap Open API to acquire business types, distributions, and other data in the path space. The road network accessibility and other information on Sanhe Ancient Town were collected based on sDNA. Finally, the path space quality at the human-oriented scale was quantized from multiple dimensions. Several sampling sites were measured in situ using a thermal comfort degree tester (Jt-IAQ-503) to acquire physical environment information. Physical environment information for the rest of the sampling points was collected through Kriging spatial interposition. Moreover, video recordings of tourists in the path spaces were obtained by the researchers using a small portable video recorder in order to collect tourist behavioral data.
　　Based on the above data collection and processing, 12 types of typical path space types in Sanhe Ancient Town were summarized. The behavioral preferences of tourists were analyzed by combining the frequency and proportion of four types of tourist behaviors (sightseeing, shopping, resting, and walking). A hierarchical multi-regression model was then developed. In addition to a simple regulating effect test, the direct and indirect influences of the multi-dimensional space quality and physical environment on various types of tourist behaviors were analyzed. The relationships between the multi-dimensional space quality, physical environment, and tourist behaviors were evaluated. Further, the degree and mode of action of the different influencing factors were summarized. The results demonstrated that the four types of tourist behaviors in summer were directly or indirectly influenced by the multi-dimensional space quality and physical environment. Sightseeing, resting, and shopping behaviors were mainly affected by the multi-dimensional space quality. These behaviors were negatively influenced by the physical environment, while walking was positively influenced by physical environment factors. These findings provide suggestions and theoretical support for building pleasant street spaces in Anhui-style ancient towns similar to Sanhe Ancient Town. The findings also have significance for promoting rural revitalization and the sensible development of the tourism industry in characteristic towns in China.

Key words: ancient town, sightseeing path, space quality, physical environment, regulating effect, tourist behaviors

摘要： 探析人本尺度下古镇游览路径多维空间品质和物理环境对游客行为的综合影响机制，以三河古镇为例，利用卷积神经网络、Open CV、sDNA量化路径空间品质；通过实地测量、空间插值获取物理环境数据；实时录像获取游客行为。构建分层回归模型对多维空间品质、物理环境与游客行为进行解释，总结各因子作用程度和作用方式。研究表明：夏季游客四类行为均受到多维空间品质和物理环境的直接或间接影响；观赏、休憩和购物行为主要受多维空间品质影响，步行行为主要受物理环境影响。为古镇街道空间场景优化提供依据，推动特色小镇中古镇旅游产业健康发展。

关键词: 古镇, 游览路径, 空间品质, 物理环境, 调节作用, 游客行为