Smart cities and decarbonization of the transportation sector provoked significant environmental policy reforms, shifting the urban environment away from car-centric planning to non-mechanized transportation. Thus, walking and cycling became natural alternatives due to their zero-carbon operation and health benefits. Computer vision (CV) techniques offer promising capabilities for analyzing and improving pedestrian sidewalk infrastructure and usage. Despite the current efforts, the literature lacks a comprehensive synthesis of studies applying CV to sidewalks, casting a shadow on the utilization of further artificial-based algorithms in this domain. Hence, this study provides a state-of-the-art analysis of CV applications on sidewalks by presenting 1) a scientometric analysis of existing studies, 2) systematic discussions of CV applications and architectures, 3) proposal of a framework for applying CV in sidewalk studies, and 4) existing gaps and future research opportunities. The publication trend analysis indicated accelerated growth in sidewalk-focused CV research since 2018, with peaks in the last five years, indicating rising scholarly interest. The keyword analysis revealed seven clusters, where “pavements”, “computer vision”, and “deep learning” are the top keywords based on their degree centrality. The systematic analysis identified and discussed CV applications to sidewalk accessibility, path conditions, and user behaviors, showing that convolutional neural networks are the most adopted architecture in this domain. This study offers academics and industry professionals essential perspectives into the developing research area of leveraging CV for analyzing diverse facets of sidewalks. It highlights the imperative to tackle identified research gaps via innovative techniques and collaborative efforts, in addition to openly sharing datasets and source code.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号