Predictive Analysis of Accidents and Weather 

 

Motor Vehicle Accidents per Emergency Call Data

One of the many potential hazards facing modern drivers is the variability of weather. In order to combat this hazard, we in the SCAL research group are conducting a project to find the correlations between injuries sustained from vehicular accidents and weather occurrences. After receiving 911 accident call reports from the Hamilton County Emergency Communications District, we proceeded to gather weather data from local weather stations in Hamilton County to discover weather conditions pertaining to each call report. These reports included variables such as temperature, weekday, the weather condition during the accident, and hour. Currently, the team is working on implementing daily received 911 Accident Reports into our study to build a more robust prediction system. By creating a predictive model, we will be able to inform drivers of the current weather conditions and give them a preemptive warning of certain weather occurrences that may yield to a higher chance of a damaging accident occurring.

 


 

3rd Eye Application

 

Predictive Routing Algorithm Based on Historical and Real-Time Data

This project uses the combination of real-time and historical data (through historical 911 emergency data (accidents, road conditions, etc.) and real-time data through computer vision) to create a dynamic mapping application that can be used as an early warning and rerouting system. This application will be able to analyze real-time and historical traffic conditions based on time of day, location, and weather conditions to identify the most optimal route.

 

Historical Data Sources

 

  •      911 Accident Data: Collecting specific 911 traffic calls, which contain locational and temporal information about the traffic accident, allowed for the identification of specific weather occurrences and roadway geometrics present during the traffic accident.  By utilizing this procedure, we are able to study and analyze the effects that the weather and roadway geometrics have on the traffic accidents. In our previous research, we performed a similar task. Courtesy of the Hamilton County Emergency Communications District, the research team received 911 vehicular accident related calls daily. These call records initially include the physical address of the accident, the citythe accident occurred within,  the latitude / longitudecoordinates, level of injury severity, as well as the time the accident was reportedand the time it was resolved.

911 Accident Data Example

911 Accident Data spreadsheet

 

  •      Dark Sky Time and Location Specific Weather Data: After obtaining 911 data, we use the locational and temporal information to determine what the weather conditions were like during the 911 accident.  The specific weather conditions we include are Event, Conditions, Temperature, Maximum Temperature, Minimum Temperature, Dewpoint, Humidity, Visibility, Cloud Coverage, Precipitation Type, Precipitation Intensity, Precipitation Intensity Max, Precipitation Intensity Time, Event Before, and Condition Before.  All of these variables have been analyzed and studied to find their individual correlations to accident frequency.

DarkSky weather data

Dark Sky Weather Data Spreadsheet

 

  •      E-TRIMS Road Geometric and Traffic Data: By connecting specific 911 accidents to specific roadways, we can study the particular road geometrics that have correlations to traffic accidents occurring, such as presence of horizontal curves, shoulder/lane/median widths, speed limit of road segment, rural/urban classification, coefficient of pavement friction, pavement mix design, road surface, shoulder type, number of lanes, and more.

E-Trims Roadway Geometry

E-TRIMS Road Geometric and Traffic Data

 

Real-Time Data Source

 

Real-time data will be gathered by the infrastructure cameras on the new MLK Corridor expansion. A graphical representation of the new corridor can be seen below. This corridor will provide real-time date collected by a variety of sensors such as LiDAR, cameras, air quality sensors and more. Our current focus is real-time data through computer vision. The infrastructure cameras run a computer vision algorithm to detect and determine approximate geo-coordinates of the object in relation to the camera. Along with absolute geo-coordinates determined by mobile application users, these coordinates can be used to give the routing algorithm of real-time traffic flow, potentially accidents, or road debris located on the user's current route or be used to create a more efficient route for the user given the new, real-time information. A graphic of this infrastructure camera architecture can be seen below.

 

Graphical Representation of the MLK Corridor Expansion

Graphical Representation of the MLK Corridor Expansion

 

Computer Vision Infrastructure Camera Architecture

Infrastructure Camera Architecture from Camera to Application