How do you use drones to speed up accident forensics?

Road accidents often require officers and their colleagues to document the accident scene. This critical step, also known as accident reconstruction, is a delicate balancing act between accuracy and speed. After all, every second the road remains blocked means more traffic and an increased safety risk for everyone involved.

Drones are a perfect tool here. Not only to help document the accident site but also to keep public safety officials safe. sUAS mapping takes a fraction of the time of traditional or older documentation options. This not only results in quick clearance. Actual accident locations are kept in 3D with detailed information and centimetre accuracy.

In this article, we talk about this workflow. In general, several methods are used. Test it out and find what works best for you that way.

Summary

1. Collecting data
2. 1. Enable appropriate camera settings and RTK.
   2. Schedule an automated mapping or oblique mission in DJI Pilot at ~75-125 feet.
   3. If necessary, fly a circle around the location at ~25 and 50 feet with timed photo settings enabled.
   4. If flying at night, consider motion blur and camera settings.
3. Processing data in DJI Terra
4. Using data in 2D maps and 3D models/point clouds

1. Collecting data

Accident site considerations

Before capturing data, consider whether accuracy is an important factor or, whether it will only be used for situational awareness. Do you want to increase accuracy? Then using ground control points and connecting the drone to RTK are appropriate. Control points, in turn, are a way to verify accuracy. Recording CPs and GCPs can be completed with a survey-grade GNSS RTK rover.

Not using control points? Then it is possible to estimate relative accuracy from site measurements or by looking at the georeferenced RMSE in the DJI Terra quality report.

While capturing data, also keep in mind that propellers and other evidence may be washed away and disturbed.

Data capture settings

With a 0.7s photo interval, 15 m/s maximum speed, a mechanical shutter and 4/3-inch CMOS Wide Camera, the Mavic 3 Enterprise is a great choice for accident reconstruction. This drone can also be linked with an RTK module for corrections. This allows only a few control points to be used instead of many. For ground control, you can use DJI Pilot 2, a free software that is automatically installed on the remote control and updated with that same remote control.

We recommend using the data capture settings below:

• Image format: JPG
• Aspect ratio: 4:3
• Mechanical shutter: on
• Focus: off
• RTK: on (recommended)

Automated data collection

As a first step, automatic data collection is important to collect overlapping photos of the whole scene. During the day, you obtain this via the flight route section of the Pilot 2 app, where you choose Mapping, Mapping with Smart Oblique, or Oblique.

In mission planning, we strongly recommend including the parameters/settings below.

• Default overlap of at least 70% side and 80% front. Consider a higher overlap if the subject does not show much variation so that the photogrammetry software can merge the images.
• When performing a mapping mission with the M3E or P1, select the smart oblique option if the area allows it.
• For a smaller accident site, it is generally advantageous to fly at the lowest altitude that is safe to avoid all obstacles. This is often an altitude between 75 and 125 feet. As the accident site gets larger, you need to evaluate flight time and resolution, which are inversely proportional to the required photo overlap.

Manual data capture

After the high-level data has been captured, it is important to consider whether it is necessary to take additional images of the site at a lower altitude. For example, if there was a tree canopy on either side of the road, and the original automated flight was conducted at 125 feet above sea level, a manual flight at lower altitude could collect additional images.

You achieve manual data collection most easily by:

• Taking photos while descending from the higher altitude for data capture.
• Adjusting the drone’s position and camera angle to place the vehicle or area of interest in the centre of the frame. While doing so, monitor the vertical angle to keep the sky and surrounding landscape out of the picture.
• Enabling the timed photo setting (i.e., every two seconds) to collect overlapping photos while flying.
• Flying a circle around the scene at ~50 – 75 feet. Descend another 25 feet to ~25 – 50 feet. Fly another circle around the scene at ~25 – 50 feet.
• Stopping the timed shots and landing the drone.

Night accident reconstruction considerations

Because the drone moves at a high-speed during data collection, the above processes work best during the day. If the drone needs to collect data at night, there are other considerations.

First, the pilot must be very careful regarding flight safety, as the obstacle detection vision systems are disabled in low light. Therefore, consider engaging a visual observer. He can closely monitor the drone’s status during flight.

Next, the accident site should be well lit. Set up lights at a height (8 feet) around the site and point them downwards. This will create a halo of light around the subject. At this point, the operator can fly the drone over the accident site to see if the lights create an overexposed photo. If so, adjust the light power or reset the lights.

If you cannot see details in the images, they will not come across in the maps or models. In general, it is easier to work with some shadows than overexposed images.

However, an accident site may be larger than a single area you can illuminate with the available light. Here, placing ground control points at the edges of the illuminated area may be an option. These can then be included in multiple data shot.

Automated missions at night

When planning an automated mission at night, you will want to significantly reduce the flight speed to the lowest possible value. This way, you avoid blurriness and plan the flight only in the wreath of light created by the lights. It can also be advantageous to fly at 40 feet when there are no obstacles. Which would generally be the simplest approach for data collection. However, keep in mind that the photos will be checked after this, and the drone will be closely monitored during flight.

Manual data collection at night

Want good image quality? Then you may need to adjust the camera parameters manually instead of letting the automatic setting do the work. For example, you can set the shutter speed higher if parts of the location seem blown out and you can’t move the lights.

Extra tip: a longer exposure value lets more light through. However, depending on the load and light being used, the pilot may need to pause while the photo is taken. This way you avoid blur. In general, a shutter speed of 1/60 will help prevent blur even if the drone is moving. While capturing data manually, the grid on the screen in the Pilot 2 app can be used as a reference for overlap.

Another option often used is to take a video while flying and extract the frames later as images.

2. Processing data

Processing imagery in DJI Terra can produce an accurate 2D map and 3D models to document the location. The process is quite simple.

• Import images and choose settings/output files
  • Are you using the RTK module with the Mavic 3 Enterprise? Then the metadata will automatically take corrections into account. In contrast, when using an NTRIP service, it is important to define the correct coordinate system for the images before processing. Specifically, NAD83 EPSG 4269.
  • If too large an area is captured, after importing the images, you can select a particular group to process based on the x/y GPS location. You can also select a particular area of the location to process after applying aerotriangulation.
  • Full aerotriangulation
  • Add GCPs/control points
  • Process 2D map and 3D model: a powerful Windows computer is required here.
  • After processing the images, you can check the report to see the accuracy of the checkpoints and measure the scale constraint.

We encourage you to try out DJI Terra with a 1-month trial.

3. Using output

After you have processed the data, there are several options for using it. In general, this allows trial investigators, prosecutors, and jurors to view the accident scene at any time and from any angle.

For easy sharing, you can integrate screenshots of the 2D map or 3D model into a PDF file. Or you can turn on automatic 360 rotation in DJI Terra to integrate a screenshot into your output model.

The files themselves – including images, GCPs, 2D map and 3D model –can also be shared, can be viewed on another Windows computer on which DJI Terra (free for viewing) or other GIS software is installed.

By importing the data into specialized software, you can additionally determine the bullet trajectory and line of sight or perform crush, impulse, and other forensic analysis. The point cloud can even be merged with data from ground scanners thanks to free software such as CloudCompare.

Want to know more about using drones in accident reconstruction?Contact us!