Analyzing Children’s Behaviors Based on AI Recognition Approach to Promote Child-Friendly School-to-Home Street Design

School-to-home streets are an important part of the public space network of child-friendly cities. The improvement of their child-friendliness needs to be based on the investigation and evaluation of the current usage status. The traditional site investigation methods are limited by manpower and time, lack of efficiency and effectiveness. This paper proposes a new analysis framework for children’s street behaviors based on AI recognition methods, including action recognition algorithm, human flow statistic algorithm, and target tracking algorithm. It takes one school-to-home street in the Mingdong community, Ningbo city as a case study. Based on the AI recognition of this street’s surveillance videos, the result data verifies the effectiveness of street usage analysis. This is valuable to promote the child-friendly school-to-home street design in the future.