Research on Small Public Space Recognition Based on Street View Images: A Case Study of Beijing Beixiaguan Street

Abstract

Abstract As urbanization continues, people are increasingly seeking spiritual fulfillment beyond material needs. The demand for public space is growing, but the quantity is inadequate, and some spaces lack the required quality. In response, attention is shifting towards underutilized small public spaces in cities. These spaces are gaining importance as a means to address spatial and social challenges in declining inner cities through renovation or addition. Traditional methods of locating small public spaces require a lot of human and material resources for on-site investigation, which is costly and inefficient, while street view images, as a new data source, contain environmental information and spatial information around almost all roads in the city, thus becoming an option for identifying the location of small public spaces. This paper integrates three algorithms of image similarity matching, image target detection recognition and image semantic segmentation, labels and trains target element datasets, and implements a target detection recognition method based on YOLOv5 algorithm to identify streetscape images containing target elements, and takes Beijing Beixiaguan street as an example to comprehensively and efficiently identify streetscape image information of Beixiaguan street and accurately identify small public spaces in Beixiaguan street. The study shows that the method of identifying small public spaces based on streetscape images provides an important solution for efficiently discovering and utilizing small public spaces hidden in urban corners, which can promote the efficiency of urban management planning and contribute to the vision of human-centered urban renewal.