UAS Crash Investigation Performance Analysis

On Scene Crash Investigation Time

Figure 1 shows the average elaspsed time for on scene crash invesigtaion for the Lake County Major Crash Assistance team for the period from 2012-2016.  The chart also shows the average elapsed time required for the forensic mapping tasks that are performed on site.  It is apparent from the charts that forensic mapping consistently consumes more than 80% of the total on scene time.  The mapping process using traditional engineering survey stations is the longest duration task on the critical path to reopening the roadway.


Historical On Scene Crash Investigation Time
Figure 1: Historical On Scene Crash Investigation Times 2012-2016

During late 2015 and 2016, the Lake County Major Crash Assistance Team evaluated the role that sUAS can play in reducing the time required to complete the forensic mapping of crash scenes by conducting the mapping using both traditional engineering station and sUAS photogrammetric techniques in parallel. Figure 2 shows the time required using each technique.


Figure 2: Comparison of Forensic Mapping Times using Total Station and sUAS Photogrammetry

The average sUAS based data collection time was 23 minutes and the average Total Station data collection time was 171 minutes.  On average the use of sUAS alone would have reduced the on site scene invesigation and road closure times by 148 minutes or two hours and twenty-eight minutes.

It is worth noting that the first scene mapped using sUAS took significantly longer due to an abunduance of caution in our data capture process.  Subsequent refinement of the altitude, location, angles and frequency of image capture allowed the total capture time to be reduced significantly.  The other scene that took over fifty minutes for sUAS capture was complicated by overhanging trees which required photography both above and below the canopy, a vehicle that was surrounded by trees and a  scene that spanned more than 700 feet of roadway.

Photogrammetric Accuracy

Several techniques have been used to validate the accuracy of linear measurements achieved using sUAS.  We have compared against measurements taken on the ground with tape, against Google Earth and against data from the GIS division of Lake County. These comparisons show an accuracy for the photogrammetric models of +/-0.3%.  We plan to prepare a more formal statistical analysis of the accuracy of our results against the measurements from the Total Station, but at present this level of accuracy is more than sufficient for our needs, so further validation has not been a high priority.