Chun Xu

 

In the urban environment, street comfort contributes to the user preference on public space. The interaction of different spatial components determines how comfortable the streets could be. The proportional relationship of spatial elements gives various spatial characteristics to public space including streets, and these elements and characteristics will, to a certain extent, have effects on the quality of public space - further to promote or discourage the use of public space for particular outdoor activities.

 

In this study, we used movement heat map data and population size distributions to fit the present street activity status to the typical active and idle streets. The Google Street View images were used to assess the active space, especially the scale of selected spatial elements, which were then contrasted with the motion heat maps to examine the value of each spatial feature. The proportional relationship of spatial elements in street space was determined using computer vision techniques like semantic segmentation and Mask RCNN feature detection. A large sample of data produced the effect of spatial elements on street activity, which led to the conclusion that spatial elements are related to people's perceptions of different streets' outdoor activity relationship between preferences.

 

Further research is expected to analyze more sample data to continuously train the model, and to use machine learning to obtain models with a universal evaluation system that not only gives smart recommendations for streets that are suitable for outdoor sports, but also for streets that are less active in the space. The goal is to apply analytical results in renovation and enhancement design, in order to build a healthier city in the perspective of user experience.

​