The effects of urban greenway environment on recreational activities in tropical high-density Singapore: A computer vision approach

The urban greenway has been increasingly recognised as an important type of green infrastructure especially for land-scarce, densely-populated cities to efficiently provide their residents with continuous public spaces close to nature for recreation. Nevertheless, empirical studies on urban greenways and their recreational use rarely focus on high-density environment. Moreover, most research endeavours in this field are also largely confined to the subtropical climate, whereas much of the world's future urban growth is projected to occur in the form of high-density mega-cities in much of tropical South and Southeast Asia. In view of these gaps, this study proposes a new approach that employs Computer Vision tools to examine the effects of the greenway's physical environment on recreational activities, taking tropical Singapore as the test bed. The semantic segmentation model, PSPNet and the action detection model, ACAM are adapted and applied in conjunction with geographical information system tools to measure the greenway's physical environment and people's recreational activity at the human scale, and analyse their relationships. The result reveals a pattern that sees the clustering of different types of recreational activities at different time periods. It also reveals the relationships between recreational activities and specific environmental features, which were observed to have influenced the overall spatial distributions of the recreational activities. The finding also corroborates the design strategies for Singapore's future urban greenways and offers a reference for engaging community groups to participate in the maintenance of urban greenways.