Documenting Vehicle Crashes with Drone Photogrammetry
Written by Krista Montgomery   

It was an early summer morning in June when a minibus from the Providence Center, an organization serving those with mental illness, headed south on Ritchie Highway in Maryland. The morning was already warm as the vehicle passed through wooded areas heavy with greenery. At precisely 7:15 a.m., a random and unfortunate event struck the moving bus, in the form of a falling tree.


Aerial image from drone of an accident site.

Responding firefighters worked for nearly 35 minutes to free the seriously injured 57-year-old driver, who was immediately transported via police helicopter to Shock Trauma. Two other passengers with minor injuries were taken to a nearby medical center, and the southbound route from Chautaugua Road to Route 50 closed for an hour.

“Police stated after the investigation that the accident was an ‘act of nature’ and that it was unlikely that the driver could have avoided it,” said Cpl. Greg Russell of the Anne Arundel County Police Department, who supplied the details of this case.

During that hour, county police used a small quadcopter drone, called a DJI Phantom 3, to capture videos and images of the scene. As with any other accident scene, it was in everyone’s best interest to clear the area as soon as possible, which couldn’t be done until scene had been documented. Accident responders have a myriad of methods to document sites: laser scanners, total stations, and photography (or a mixture of the three) are used to gather measureable data and produce a 3D point cloud reconstruction. These methods, however, can take a large amount of time and require personnel trained in surveying, who may not always be on hand when an accident occurs.

More and more, law enforcement is looking to drone use for accident reconstruction and documentation. Used in tandem with photogrammetry software, the small flying robots have huge potential for quick and accurate documentation of vehicle accident scenes: especially pertinent when there is a large loss of property or life.

“When you look at the conventional methods, you’re looking at time,” said James Addison, a forensic reconstructionist in the private sector and retired police lieutenant. “You’re under a lot of pressure. To have a freeway closed for five or six hours to investigate a scene creates a lot of tension.”

Drones can easily tackle larger distances for faster coverage: the larger the accident area, the better the efficiency of a drone to collect the data. In a study by Unmanned Experts, Drew Jurkofsky, a Colorado police officer and accident reconstructionist with nearly 20 years of experience, found that using a drone for accident documentation instead of traditional methods cut time spent in the roadway to nearly one fifth, and measuring time to around one third.

For the minibus accident in Maryland, Anne Arundel County police used the 154 overlapping images taken by the DJI Phantom 3 drone and photogrammetry software, Pix4Dmapper Pro, to reconstruct the scene. The resulting point cloud, 3D model, and orthomosaic outputs were implemented in CAD software to make diagrams and conduct a follow up analysis.


Post-Impact Movement Track example.


Damage Profile and Crush Measurement example.


3D point cloud of the accident scene.


Accident scene reconstructed using drone images.

For accident cases similar to this one, the road could have been shut down for up to three hours (using LiDAR to document) instead of the average 20 minutes taken by a drone. It is worth mentioning that closing roads or reducing traffic flow in order to document a site is not only costly to responders and drivers in terms of time and finance, but also puts lives in danger. According to the Federal Highway Administration (FHWA), every minute an accident continues to be a hazard increases the chance of additional accidents by 2.8 percent.

Traffic related fatalities are consistently a leading cause of officer death, and this risk is reduced both by lowering the amount of time law enforcement and reconstructionists spend gathering data on site and their exposure to traffic.

By deploying drones as a tool to clean up accidents more quickly, said Steve McKinzie, CEO of an accident investigation and reconstruction firm, “We cut the economic loss of communities where critical highway incidences occur, we improve the safety of the working environment of the first responders, policemen, firefighters, and EMS, and we reduce the secondary collisions.”

For the Anne Arundel County police, key elements for the minibus analysis include the diagrams of the damaged vehicle profile superimposed over the undamaged one, actual measurement of the damage, and the vehicle’s post-impact travel. Here we provide sample diagrams to explain those elements in more detail. (Note: Due to regulation, these diagrams are not directly related to this accident. They are displayed as samples to showcase the report elements.)

Damage Profile and Crush Measurement

A diagram of the damaged vehicle profile can be generated in CAD software by importing an image orthomosaic created by Pix4Dmapper software. This diagram is then superimposed on an undamaged diagram in the database. Vehicle damages are then measured and quantified to give information on the speed the vehicle was traveling and the force it endured during the crash.

Post-Impact Movement Track

Along with the simulation and the crush force information, the trajectory of the vehicles involved in the accident is documented. This gives the information on the vehicle speed and how it moved, in terms of distance and rotation angles, after the crash happened.

The goal of data collection by drone is to produce a reliable diagram, something that can serve as evidence in case it needs to be shown in court. Ultimately, the cause of the accident needs to be determined. Image data collected in just one drone flight can be used by photogrammetry software like Pix4Dmapper Pro to produce a 3D point cloud, model, measurable orthomosaic, and detailed report. In this specific case, the accident area of 2.26 acres was mapped with a ground surface distance (GSD), or distance between pixel centers measured on the ground, of 1.6 centimeters.


Making measurements on an accident scene reconstruction in Pix4Dmapper Pro software’s rayCloud editor.


3D model of an accident training scene, created from images in Pix4Dmapper Pro software.


Making measurements on an accident scene reconstruction in Pix4Dmapper Pro software.

“You have a two-fold purpose with UAVs,” said James Addison. “You can get the point cloud data and use it to take measurements and recreate the scene in 3D. Then, you still have the photos to present as evidence themselves, or to render into a scaled, aerial view orthomosaic as a diagram.”

Even more important than producing a model that helps describe an accident scene to a jury, is having measurable data: the unofficial and approximate accuracy standard is around 2–5 cm, though this varies depending on case needs.

“From a scientific process, you’re not so worried about visualizing the data, as you are being able to analyze the data,” said Steve McKinzie, who emphasized the importance of being able to measure distance and angles of a site. “We always come back to this issue of how and why an incidence occurred. And if we can’t answer the how and the why it occurred, we can never prevent a similar instance from occurring.”

With increasingly automated drone flight planners, such as Pix4Dcapture, and photogrammetry software, less specialization is needed to accurately document an accident scene. Not only is the workflow less invasive, it is ideal in situations with extreme time, personnel, or equipment restrictions.

Drone image acquisition does not have the same sensitivity to obstacle occlusions or moving objects that laser scanning and total station methods do, making a complete and clean model easier to acquire. Reconstructing imagery in software like Pix4Dmapper is not camera restrictive either, meaning any camera on hand can be used to collect data from an accident. It’s even possible to use video from a smartphone, which the software uses to extract still frames.

In the case of the minibus accident in Providence, the Anne Arundel County Police team was able to quickly gather image documentation of the minibus accident, saving details of both the damaged vehicle and the surrounding site. By using the overlapping drone images, they recreated a detailed 3D scene in Pix4Dmapper Pro software that they were able to revisit again and again for different types of analysis.

Far from being an unsafe, invasive, or violent tool in the hands of law enforcement, drones and photogrammetry software are likely to become a standard part of the toolkit for surveying accident scenes. Their use has proven benefits and potential to revolutionize the way we deal with vehicle crashes.


About the Author

This e-mail address is being protected from spam bots, you need JavaScript enabled to view it is a writer and content manager at Pix4D, a rapidly expanding developer of photogrammetry software solutions. Based in Switzerland, the company aims to democratize drone mapping and to provide end-to-end solutions for its growing professional user base.

Special thanks to Cpl. Greg Russell and Lisa Chen.








 

 
< Prev   Next >






Item of Interest

The language barrier between English-speaking investigators and Spanish-speaking witnesses is a growing problem. (Updated 28 February 2011)

Read more...