Objective For urban street space, a main place for residents’ daily public activities, its greening structure plays a crucial role in influencing walking index, environmental assessment, residents’ health and economic benefits. Street greening can significantly enhance residents’ life satisfaction, especially in high-density urban environments. With the development of computer technology, it has become possible to combine visual analysis techniques to conduct a fine-grained research on urban street greening structure. Studying the spatial and temporal evolution of street greening structure not only helps understand the spatial experience and landscape changes of urban streets, but also has great significance in scientifically evaluating and optimizing urban street greening, and promoting the high-quality and sustainable development of urban street space. Urban street greening structure is a three-dimensional plant composition concept, which usually includes high-level trees, mid-level shrubs and herbaceous plants. Through historical research and analysis of street view images, the research shows that traditional street greening measurements cannot effectively reflect the three-dimensional visual experience of residents, and therefore new methods are needed to more accurately analyze the greening structure of urban streets. In summary, the purpose of this research is to propose a new research framework to finely analyze the change trends and influence mechanisms at the level of urban street greening structure. Taking the Chinese city Shanghai as an example, the research adopts advanced artificial intelligence technology and time series street view data to conduct an in-depth research on the greening structure of urban streets.

Methods At the data level, the road network is downloaded through OpenStreetMap, and street view sampling points are set at an interval of 50 m along the road network. Time series street view image data on the central urban area of Shanghai in summers during the period from 2013 and 2019 are collected via Baidu Map. At the model level, a self-annotated and trained deep learning model is used for semantic segmentation to identify green structures (trees, shrubs, and herbaceous plants) in the street view images. Combining multiple data sources, the DeepLabV3+ model is trained and migration learning is performed to improve accuracy and generalization. At the analysis level, statistical methods are used to classify and analyze the green structures in the street view images, taking into account the frequency and distribution of the green structures. The greening structures of Shanghai streets are classified into five categories based on the type of greening, and the spatial distribution and changes of these categories are analyzed.

Results From 2013 to 2019, the urban street greening structure in the central urban area of Shanghai changed significantly. The results of the research show that the proportions of trees, shrubs and herbaceous plants increased by 25.09%, 19.32% and 42.39% respectively during the period from 2013 to 2019. This increase reflects the enrichment of the urban greening structure in terms of plant species and distribution. In particular, the comprehensive greening area (cluster 1) increased by 23.99% in 2019 compared to 2013, and the change was particularly significant in Pudong and Yangpu districts. The cluster analysis of the greening structure of urban streets shows that the spatial distribution and quantity of each type of greening structure also changed from 2013 to 2019. For example, both tree-dominated greening areas (cluster 4) and herb-dominated areas (cluster 3) show an increase in quantity, suggesting that urban planning has made progress in promoting diverse greening and improving ecological quality. In terms of the spatial distribution of the city, comprehensive greening areas in most administrative districts increased in 2019, reflecting the city’s continued investment in enhancing the living environment for residents and improving ecological functions. Especially in Pudong New Area, the vegetation changes were particularly significant due to the late development of the area, showing the positive adjustment of urban planning and greening strategies.

Conclusion In conclusion, it is in line with the requirements of the urban construction era to study the evolution and optimization strategies of urban street greening structures. At the level of policy and greening impact, Shanghai has promoted the quality of urban greening through the implementation of policies such as the Shanghai Greening Regulations. The policies have not only enhanced the aesthetics and functionality of green spaces, but also protected them through legal provisions and promoted community participation and environmental protection awareness. At the level of socio-economic benefits of greening structures, street greening can indirectly enhance property values by providing recreational space and improving the urban microclimate, and reduce public health expenditures by increasing biodiversity and improving air quality. At the level of classification and application of street greening structures, different street types require different greening structures to meet their functional needs according to their user groups and environmental characteristics. The dimension of greening structures and urban ecosystems emphasizes that through rational design of greening structures, the ecological system of the city can be enhanced, the microclimate improved, the biodiversity increased and the ecological network formed. This research can provide urban planners and policy makers with a new perspective on how to improve the quality of urban environment and residents’ life through urban street greening. It also points out the challenges in the process of urban greening, such as financial constraints and technical difficulties, and suggests appropriate solutions.