 |  |  |  |
Developing SafeSpot: A New Crowdsourcing Tool for Real-Time Near-Miss Detection and Risk Mapping
Principal Investigator:
Tanmoy Bhowmik, Portland State University
Co-Investigators:
- Primal Pappachan, Portland State University
Summary:
Transportation safety analysis has traditionally relied on police-reported crashes to identify dangerous locations. This reactive approach overlooks the many near-miss incidents and perceived hazards that occur daily but go unrecorded. Understanding such incidents is essential for proactive transportation safety planning, allowing for interventions before injuries or fatalities occur. However, cur...
Transportation safety analysis has traditionally relied on police-reported crashes to identify dangerous locations. This reactive approach overlooks the many near-miss incidents and perceived hazards that occur daily but go unrecorded. Understanding such incidents is essential for proactive transportation safety planning, allowing for interventions before injuries or fatalities occur. However, current practice lacks a systematic way to collect and analyze near-miss data, creating a critical gap in our ability to prevent crashes. This research proposes to address this gap by developing a crowdsourcing platform, called SafeSpot, with an interactive map that will allow individuals to submit reports of near-misses and perceived driving hazards. Users will be able to input details about their location and nature of the incident/hazard such as sudden braking, poor visibility, aggressive driving, or other unsafe conditions. As these crowdsourced reports may contain personally identifying information (e.g., precise user location), we will employ a secure and privacy preserving architecture ensuring that only anonymized information is transmitted from the user device. The platform will be piloted in a small-scale urban area (e.g., Portland, OR) to assess usability, participation, and data coverage, with flexibility to scale to broader communities. Reported data will then be used to train predictive models that estimate the likelihood of near-miss incidents based on roadway design, land use, traffic exposure, and temporal factors. These models will then be applied to the broader urban network to generate a synthetic dataset of near-miss events, enabling spatially continuous risk estimation even in areas with limited user reporting. This synthetic data will be integrated with historical crash records and built environment characteristics to assess alignment between predicted near-miss hotspots and actual crash patterns. These models will then be used to generate a synthetic dataset of near-miss events across the wider network, enabling spatially continuous risk estimation even in areas with low user engagement. By doing so, SafeSpot offers a scalable alternative to expensive sensor-based or proprietary data sources. Finally, this unified dataset will be integrated with historical crash records and street design features to uncover emerging safety risks and latent hotspots. These insights will inform transportation agencies’ proactive planning and engineering countermeasures, guiding interventions such as street design changes or policy measures before severe collisions happen.
See More
Project Details
| Project Type: | Research |
|---|---|
| Project Status: | In Progress |
| End Date: | August 15, 2027 |
| UTC Funding: | $69,387 |