CRJ Key Network Partner, Pix4D’s Caroline Bailey and Angad Singh discuss using drones and aerial mapping for crash investigation and other public safety projects with Aerial Metrics co-founders, Iain Lopata and Stan Taylor.
Using drones and Pix4D software for public safety is beneficial to society. The time taken to document crash investigation evidence is reduced dramatically and it can provide compelling support in the courts. Image: Andrey Suslov/123rf
In the small hours of a January morning, a speeding vehicle veered off a major road and struck a streetlight, killing the passenger. The posted speed limit was 30 mph. The responding officer took 61 photos with a Nikon DSLR camera, including 18 that captured the skid mark left by the tyres on the roadway. At the time, there was insufficient evidence to determine the speed of the vehicle.
Almost a year later, Aerial Metrics was asked to fly a small unmanned aircraft system (sUAS) mission at the scene and combine the photos taken from the drone with the original DSLR photography in order to build a model with Pix4Dmapper software. From the model, the Aerial Metrics team was able to measure the length and curvature of the skid mark and determine a range of possible speeds for the vehicle, which were in excess of 60 mph. This evidence allowed charges to be filed against the driver, which would not have been possible without this technology.
The founders of Aerial Metrics say: “We wanted to explore the potential for the use of sUAS to improve the process of crash investigation. So, together with the Lake County major crash assistance team (MCAT), we developed a system that employs sUAS and photogrammetry to reduce road closure times, improve officer safety, and capture better evidence at crash scenes.”
When asked about the workflow in the field, the Aerial Metrics team gives a breakdown: “This workflow has been developed and optimised by Aerial Metrics after years in the field to fit collision reconstruction scenarios, account for real-world obstacles like trees and ensure more than enough data has been captured before scene clean-up:
- Once the scene is safe, evidence on the roadway and surrounding areas is identified and marked with spray paint, evidence tents, or versa-cones;
- We use two or more scale reference measurements taken between easily identifiable points at least 50 feet apart. These are used later to verify the accuracy of the photogrammetric model;
- Before we start with mapping flights, we do a quick pre-flight check to ensure hardware/software function and regulatory compliance;
- Three flights are conducted over the scene or subsets at progressively lower heights. The pilots are trained to determine the appropriate heights and photographic overlap distances for each layer. The first and second layers cover the same area. The third (lowest) covers only the area of key evidence and may also include obliques of vehicles. The easiest way to think about it is that the middle run is primary, the higher run is insurance in case of inadequate overlap, and the lowest run improves resolution/GSD for the most critical evidence.
All images are processed in the same project. The flight heights are selected to ensure that the three layers will calibrate together. This workflow reduces total flight time while still ensuring very high-resolution models for capturing even small evidence details such as shell casings.
Flight 1 - Drone passes down the middle of the scene to capture its entire width with 75 per cent frontal overlap
Flight 2 - Drone is flown lower, at two-thirds of the height of the first flight. Includes a pass down both the left and right-hand sides of the scene
Flight 3 - Drone is flown 50 per cent lower again if there is a need to identify specific evidence; and
- Once the flights are concluded, the evidence and vehicles can be removed and the roadway opened for traffic.”
Benefits to society
The teams then discuss whether using drones and Pix4D software for public safety is beneficial to society. The resounding answer from the Aerial Metrics team is: “Most definitely. The time taken to document crash investigation evidence has been reduced from two to three hours to only 15-20 minutes.”
The team explains that this can lead to roads being opened sooner and improved officer safety, with a reduction in demand on their time. It can also reduce the risk of secondary crashes owing to stopped or slowed traffic and reduced fuel consumption from extended traffic delays.
The main benefits of using the technology are enabling richer data capture; being able to create a full, three-dimensional model reconstruction with millions of data points;. and achieving at least equal or better accuracy. In addition, say the Aerial Metrics team, there is the ability for retroactive measurement – with total station capture, the distances to be measured must be selected at the scene, but with sUAS capture, any measurements can be made once the model has been created.
Finally, the team says that 3D colour models can provide compelling evidence in court. Viewpoints can be changed to any location within the scene, showing what a driver or witness would have been able to see. Three dimensional models also tend to be more persuasive than 2D line drawings or primitive animations.
Pix4Dmapper is recommended for the forensic mapping of crash and crime scenes and because it is the most widely used software for this purpose in the public safety community, there is a wealth of experience in its use. It has been rigorously evaluated and peer-reviewed papers are available from the Society of Automotive Engineers that document its accuracy. Moreover, its outputs have been widely accepted in court without challenge.
Pix4Dreact is recommended for rapid response mapping missions such as wildfires, SWAT operations, floods, and other extreme events. It is simple to use and delivers scaled maps much quicker than Pix4Dmapper, but without the level of accuracy required for forensic mapping.
Next, the teams discuss important considerations in terms of choosing hardware. Aerial Metrics say that the drone’s purpose has to be considered first. There are different types of drone for different types of operation and one drone will not be able to cover every mission without incurring considerable hardware costs.
The team goes on to say: “Secondly, new drones enter the marketplace every twelve to eighteen months, so reviewing available platforms is warranted before finalising a selection. While it’s inevitable that any purchase will quickly be superseded by newer technology, we have generally found that sUAS are serviceable for at least four years with little hardware maintenance unless they are involved in a crash landing.”
In terms of setbacks and challenges when trying to adopt and start a programme, the teams surmise that the public still has concerns about how law enforcement agencies might use (or abuse) the technology. This can hamper efforts to implement drone capabilities. Aerial Metrics sets out its key points for success: “Careful selection of mission – prioritise those missions that have a clear benefit to the public, for example opening roadways more quickly after a crash or search and rescue; clear procedures defining when, where, and how the technology will be used; and early engagement with the community and consistent messaging.”
Finally, the Pix4D team asks if the technology works successfully in the dark. Aerial Metrics replies: “Yes, crash scenes can be mapped at night using sUAS. We have mapped over one hundred scenes at night and use a variety of techniques depending on the specific circumstances and equipment available.”
In terms of lighting, Aerial Metrics say that street lighting can be sufficient when combined with high contrast paint, chalk or evidence markers, and lighting towers and squad car lights can illuminate evidence areas when they are positioned strategically. The team adds that built-in drone or camera light sources usually provide sufficient illumination at typical flight heights of 100 feet and under, but the sUAS must be kept still when the camera shutter is activated to avoid motion blur caused by the required longer exposure times.
The team also recommends flying twice – to compare results generated with night imagery with results from images captured during the day. These images should not be processed in the same project, but comparing results avoids missing evidence which is not easily identifiable in the dark.