Lights, Camera, Reconstruction

Dash cams catch a lot of crazy things on the road, but it is doubtful that there are many videos topping the one currently circulating of Kia becoming airborne upon impact with a wheel that became detached from a Chevrolet pickup. The incident occurred on a Southern California freeway on March 23, 2023, and was captured by an uninvolved Tesla’s onboard camera (the ‘camera car’).

The footage of the dramatic and unexpected turn of events that transpired on the freeway, while shocking, offers valuable insight into many potentially contributing factors, all of which can be unpacked from the video.

For instance, the Chevrolet appeared to have an aftermarket tire/wheel setup, which raises questions about the installation of the left front wheel (the wheel which eventually came off). Questions regarding the vehicle’s mechanical fitness, appropriateness of the aftermarket components, and how the wheel was secured prior to the incident are all factors that may have played a part in what appeared to be a sudden component failure.

But what about the Kia? Could the driver have done anything differently? Given the timing between the wheel detachment and impact, it’s unlikely there was much, if anything, the driver could have done. However, the Kia was in the process of passing the camera car, and if it was speeding at the time, could any excessive speed have contributed to the way this incident unfolded? In other words, if the Kia was speeding, and had it not been, would the detached wheel simply have passed in front of it, or struck it at a different orientation which would not have resulted in the vehicle becoming airborne?

A Focus on the Camera

As it relates to assessing the distances, timings and ultimately, the speeds that can be extrapolated from the video, it’s first important to understand the limitations of the camera system itself – something that simply can’t be done by analyzing the online footage alone.

While it may be tempting to rely on an internet version of the video, uploading video often compresses the original file and can alter its metadata. This process can also reencode the timing data or alter the playback speed, effectively changing the actual frame rate of the video. Unfortunately, this can result in an inaccurate understanding of event timing. Ensuring the video is correctly obtained and accurately documenting its the chain of custody is imperative for not only the analysis, but also in ensuring the video is permissible as evidence in any future potential legal actions.

Testing the camera car’s onboard camera is the most precise method in which to determine any variability that can occur in frame timing. Once determined, understanding vehicle speeds and distances becomes a task of reconstructing the video in a virtual environment by way of first understanding the position and camera parameters from the camera car itself.

A Reconstruction in 3D

In these sorts of cases, match-moving is the technique of choice in determining these parameters. This technique, though primarily used by visual effects artists, has been adopted by forensic video analysts and collision reconstruction experts to determine the camera’s movement so 3D objects can be placed relative to the moving camera. The process involves tracking various pixels, usually referencing immovable roadway objects such as dashed roadway lines, signage, etc., to iteratively solve the perspective, focal length, lens distortion, and ultimately camera position over every frame.

Once those parameters are solved, we can place 3D models of the involved vehicles in a virtual environment, orienting them to reflect their position at any given frame. The alignment process of a 3D object to reflect its true position shown in a single frame or photograph is referred to as photogrammetry: the science of extracting measurement from photographs.

Figure 1: 3D placed Kia, airborne, and the Chevrolet, placed within the 3D environment.

Though arduous, this process can help answer those all too imperative questions of speed and avoidability from what would otherwise be classified as a freak accident. Furthermore, the 3D visualizations are not only compelling, but can also offer further insight, from different vantage points not originally available from the dashcam view alone.