Skip to content

Photogrammetry as a Tool for the Forensic Documentation

When a road accident happens, the authorities must close the road for traffic, sometimes for hours, as they not only need to make sure safety is restored but they also need to investigate the accident scene to determine why and how an accident happened which can take hours. This not only affects people involved in the accident but also affects others since traffic congestion, de-routing, delays, causing a lot of dissatisfaction and lost working hours.

First important part of the process is that nothing “walks away” from the scene, therefore scene mapping should be captured AS FAST AS POSSIBLE, to capture everything and allow all details stay intact.

Second, you must collect AS MANY DETAILS AS POSSIBLE, data is later used by several users, including police force, and insurance companies for specific calculations that determine reasons of why an accident happened. 

The third important point is that investigators are provided with professional tools that help them to process collected data meaning to ACCURATELY, PROPERLY MEASURE AND CALCULATE THE DISTANCE between parts at the car crash scene, angle of the crash, skid marks and final rest positions, because only by accurate measurements they can properly determine reasons of why an accident happened, so that calculations can be done “offsite” later in the office.

PROJECT DESCRIPTION

The problem

Until now local forensic department had 2 options when investigating a road accident:

  • Use of the laser scanner which provides good results, but it is expensive therefore only limited number of devices are available, which causes delays and complex resource planning, transferring devices from one part of country to another
  • Traditional approach meaning, hand sketch + around 100 photographs at the scene. This method provides the documentation of the car accident site, but it does not give a good situational awareness and the sketching of the larger site can be very time consuming with the fact that the accuracy is limited.

Approach

Local Police department wanted to understand how and if photogrammetry can help them with easier and more accurate data collection, calculating crash angles, speed of cars, identifying objects and measuring and calculating distances which they use in their calculations, exclusively using photos from drones, ground photos and photos from bystanders. The Police department wanted to understand how much faster Police can remove the crash scene and relieve the traffic and also how easier, faster and more accurate calculations with photogrammetric software can be.

SOLUTION

Fast data acquisition

For absolute accuracy, the project must be scaled and or have control points defined.  3Dsurvey targets (ground control points, perpendicular to each other with the distance in-between of 3 meters) were used as a base for orientation. This technique is 5 times faster and the cost is 1/20th of the price of GNSS device, so police can handle many cases at the same time and spend minimum amount of time on a non-secure area at the same time achieving same level of orientation quality. Compared to GCP points with GNSS devices, users do not need any training to position the targets.

Scene of the accident was captured with DJI Phantom 4 Pro, a midrange and affordable drone. Since users were not experienced in flying drones, flight mission was automated by 3Dsurvey Pilot app. The circle mission mode was used to automatically capture the images with the drone. 3Dsurvey Pilot app navigated the drone around the scene on the predefined perimeters which were set prior the flight at the scene using google maps in the 3Dsurvey pilot app. To increase the level of details collected, drone was set to collect images 3 times, each time from different hight.  These images will later on create a 3d model of the accident scene.

Once images were taken, Police were able to approve the removal of crashed cars and also open the road for traffic. Complete set of pictures was then taken to the office for the mapping process. 

Project tools and specs:

Drone – DJI Phantom 4 PRO

Camera – on board, 120 images – 3x circular flight (hight 10m, 15m, 20m) 

Number of gcps – 3 – local coordinate system;

Area of interest 50 x 50m

Mapping the scene

The process in 3Dsurvey begins with import of all the photos taken from the accident scene into the software. With the bundle adjustment step the software recognizes and finds common points between images, matches them and generates a raw 3D model – “sparse point cloud”.  Now the orientation of images needs to take place.

Because 3Dsurvey ground control points were used, we only needed to hand-pick three GCP points on a single image – all other GCP points are automatically detected. Software will automatically list all the images where the ground control points are visible, achieving highest grade accuracy (up to 2 cm survey grade accuracy) in scale and measurement.

Next step is the reconstruction that creates highly detailed 3D point cloud and generates a 3D model of the scene which is a perfect data type input for Textured Full 3D model generation. For the measurements integrated CAD functions were used, which gave the forensics exact details to reconstruct the accident (angles, distances, etc). 

RESULTS

  • Image acquisition 15 minutes – on site.
  • Complete data processing 3 hours – off site, at the office.

Output:

  • 3D model in OBJ format – standard 3D format
  • 3D sketch with vector CAD lines (all details, distances, angles, measures, object relationship) export format DXF
  • Orthophoto map in JPG and KMZ ready for Google Earth (possible accidents map creation, location defined…)

BENEFITS

  • Forensic scene visualization / Immediate use of scene mapping for presentation in the court room
  • Digitalised reality of the scene site
  • Up to 7 x less time spent to gather and processing data
  • Best possible precision and accuracy in force calculations and distance measurements
  • Up to 79% lower acquisition cost compared to other modern solutions
  • Possible reuse of outputs for maps, other stakeholders

Want to know more about this topic, check out our on demad webinar.