Reliable, Routable, and Reproducible: Collection of Pedestrian Pathways at Statewide Scale

TL;DR

The results demonstrate the feasibility of yielding accurate, robust pedestrian pathway networks at the scale of an entire state and inform procedures for national-scale ADA compliance by providing pedestrian equity, safety, and accessibility, and improving urban environments for all users.

Abstract

While advances in mobility technology including autonomous vehicles and multi-modal navigation systems can improve mobility equity for people with disabilities, these technologies depend crucially on accurate, standardized, and complete pedestrian path networks. Ad hoc collection efforts lead to a data record that is sparse, unreliable, and non-interoperable. This paper presents a sociotechnical methodology to collect, manage, serve, and maintain pedestrian path data at a statewide scale. Combining the automation afforded by computer-vision approaches applied to aerial imagery and existing road network data with the quality control afforded by interactive tools, we aim to produce routable pedestrian pathways for the entire State of Washington within approximately two years. We extract paths, crossings, and curb ramps at scale from aerial imagery, integrating multi-input segmentation methods with road topology data to ensure connected, routable networks. We then organize the predictions into project regions selected for their value to the public interest, where each project region is divided into intersection-scale tasks. These tasks are assigned and tracked through an interactive tool that manages concurrency, progress, feedback, and data management. We demonstrate that our automated systems outperform state-of-the-art methods in producing routable pathway networks, which then significantly reduces the time required for human vetting. Our results demonstrate the feasibility of yielding accurate, robust pedestrian pathway networks at the scale of an entire state. This paper intends to inform procedures for national-scale ADA compliance by providing pedestrian equity, safety, and accessibility, and improving urban environments for all users.

Figures (6)

• Figure 1: Overview of the end-to-end process for inferring pedestrian path network graph: Section \ref{['sec:Prophet']} describes Prophet, the multi-input generating a pedestrian path network from incomplete information. Section \ref{['sec:cnn']} details the segmentation network using aerial and road map images to predict class labels of pixel locations in the pedestrian map. Section \ref{['sec:opt']} explains how Pedestrianfer and segmentation network information are combined for accurate graph inference. Section \ref{['sec:edits']} describes the Skeptic protocol for manual vetting and community engagement.
• Figure 2: Hypothesized graph generated with Pedestrianfer. Yellow dots represent the hypothesized nodes for the curb modes and the nodes on the sidewalk. Blue lines represent the hypothesized sidewalks. Red lines represent the hypothesized crossings. Links are part of the crossings between the curb nodes and the nodes on the sidewalk.
• Figure 3: Illustration of node location optimization, the probability of each pixel being the corner bulb is shown as a heat map in red in (a) - (c). (a) Pedestrianfer hypothesized nodes (b) Polygons formed by the hypothesized nodes in each corner (c) New set of nodes optimized with information from the segmentation network (d) Compare to Human annotation: Green is human annotation graph Red is Pedestrianfer Hypothesized graph. Blue is the optimized graph.
• Figure 4: Qualitative results on the validation set. The segmentation results of 3 different models are shown in columns 4-6. (1) Trained with the aerial satellite image branch only (2) Trained with the street map tile branch only (3) Trained with both the aerial satellite image branch and the street map image tile branch. The segmented classes are Corner bulb, Sidewalk, and Crossing. The model using both aerial satellite images and street map images outperforms models that use either input alone. Column 7 illustrates the connected pedestrian pathway graph inferred by Prophet. We discuss these samples and quantitative results in Section \ref{['sec:exp']}.
• Figure 5: TraversabilitySimilarity Scores of pathway graphs produced by three methods on three distinct test areas. Brighter colors of each polygon indicate a higher TraversabilitySimilarity, while darker colors indicate a lower TraversabilitySimilarity. Prophet can generate graphs at scale which better capture the true traversability, than the graphs genearated by the curent state-of-the-art method Tile2Net. Human editing with Prophet + Skeptic produce graphs that best capture the true traversability.