Rotterdam 15th of May 2013

Skeye BV, a Dutch company specialised in aerial measurements using Remote Piloted Aircraft (RPA’s) assisted the Belgian Local Police Department of GOAZ (Genk, As, Opglabbeek, Zutendaal) in determining the best, most reliable and cost effective method to measure an accident scene using RPA.

A big problem in Belgium and the Netherlands (the two countries with the highest number of traffic jams in Europe) is that roads are often closed for a long time when a fatal accident has taken place. In 2009 the cost of accidents in the Netherlands was 12.5 bln. Euros, i.e. 2,2% of the Gross National Product (Factsheet December 2012, Stichting Wetenschappelijk Onderzoek Verkeersveiligheid). One of the ways to reduce these costs is to minimise the time required to take measurements and survey the accident scene.

The Genk police force aims to quickly make precise measurements with RPA-imagery taken at an accident scene. All this without disturbing active personnel such as rescue workers. An RPA is especially suited for this purpose as it hovers above the accident scene without interfering in any way. In Genk the police uses an Altura AT6 v2 that is designed and constructed by the Dutch company AerialTronics from Scheveningen. Currently this RPA is used for several police activities such as monitoring crowd movements at football games and aerial surveillance during SWAT raids. The Altura AT6 is used by security services due to its excellent flying capabilities, even in strong winds. The RPA can be pre-programmed to fly a specific route and take pictures at predefined locations.

The methodology to measure from aerial images is called photogrammetry. Photogrammetry is the practice of determining geometric properties of objects from photographic images. A more sophisticated technique, called close range photogrammetry, involves estimating the three-dimensional coordinates of points on an object. These are determined by measurements made in two or more photographic images taken from different positions. Common points are automatically identified on each image. A line of sight (or ray) can be constructed from the camera location to the point on the object. It is the intersection of these rays (triangulation) that determines the three-dimensional location of a point. When two or more aerial images have sufficient overlap, the three-dimensional coordinates of each pixel can then be determined. This is done automatically using specialised photogrammetric software. At least one reference measurement is necessary for the software for determining the scale.

For photogrammetry an RPA is normally pre-programmed at the office for taking photographs in a special pattern. However, this is not practical when arriving at an accident scene since the exact location, length etc. are not known prior to arriving. This implies that the software used for processing the images should be able to handle images taken ‘manually’ and in an irregular pattern.

For this project two different software packages have been evaluated. The first is a standard photogrammetric package and the second is 3Dias as developed by Geodelta from Delft. 3Dias was especially developed for the Dutch police for measuring accident scenes with the following constraints:

  1. Measure without applying markers on any vehicle as these can be disturbed by rescue workers.
  2. Measure in a predictable, reliable, accurate and above all a verifiable way, i.e. measurements taken cannot be refuted by other experts during a trial.

The main difference between 3Dias and more standard photogrammetric software is that it not only calculates the coordinates of the tie-points and generates a rectified image, but also calculates the quality of the tie-points, determines the quality of measurements in the images, performs statistical tests to find errors in the data and validates the flight setup, photogrammetric work and camera calibration.

An accident scene was set-up that involved two cars that had supposedly collided. Several flights were made above this reconstructed accident scene at different heights ranging from 30 to 50 meters.

Accident scene drone measurement

Images were taken straight down using a 24Mp camera. The time needed to photograph the accident scene using the RPA was around 5 minutes and included the time for take off and landing.

The used camera and lens had been calibrated prior to the flights to determine the irregularities and model possible distortions. This is a requirement if accurate measurements are to be made using photogrammetry.

Ten circular discs of which the middle can be exactly determined were placed around the accident scene and their 3D coordinates were measured using a highly accurate survey grade GPS with an accuracy of around 2 centimetres. In addition another ten control points were placed inside the accident scene to check any measurements. The distances between these additional points were measured (twenty one in total) using a fiberglass reinforced survey grade measuring tape (to avoid distortion due to stretching).

photogrammetry control points on a drone accident scene

The data was processed in three different ways:

1. Method 1: Using standard photogrammetric software and providing the software one of the 21 reference measurements.
2. Method 2: Using standard photogrammetric software and providing the software the ten GPS measured control points.
3. Method 3: Using 3Dias photogrammetric software and providing the software one of the 21 reference measurements.

The distances between the additional reference points were then measured using the photogrammetric software and compared with the hand measured distances on the ground.

Photogrammetry by drone setup for measurement

The following results were obtained:

Accuracy of drone photogrammetry

The following could be concluded from these tests:

  1. Sub centimetre measurements can be achieved with photogrammetry using an RPA above an accident scene;
  2. Using an RPA an accident scene can be surveyed in a time frame of around five minutes.
  3. Imagery taken at a height of 30 metres is detailed enough to show and measure skid and yaw marks, scratches and gauges in and on the road surface.