How Speed Is Calculated From Surveillance Video: A Starting Point for Attorneys

When a collision is caught on a surveillance camera, the video often becomes the center of the case. It’s easy to see why. Everyone can watch it, and it feels objective. But a speed calculated from that video is only as good as the steps behind it, and most of those steps aren’t visible when you press play.

This is the first of three posts on how surveillance video is used in accident reconstruction. I’ll start with the basics here and build toward the research my co-authors and I published on surveillance timestamps.

Video Is a Series of Still Pictures

A video isn’t continuous motion. It’s a series of still images, called frames, shown quickly enough that your eye sees movement. It works a lot like a flipbook. The number of pictures captured each second is the frame rate. Phone video is typically recorded at 30 frames per second or more. Business surveillance systems are often set lower to save storage space, sometimes as low as 1 or 2 frames per second.

Speed Comes from Two Measurements

In forensic video analysis, every speed calculated from video comes down to the same formula as a high school physics problem: distance divided by time. That means the analyst has to answer two separate questions.
The first is where the vehicle was. To get that, the analyst matches features in the video, like lane lines and curbs, to real measurements of the scene. This process is called photogrammetry.

Matching a camera’s view to a 3D scan of the scene is how an analyst places a vehicle in real-world space.

The second is when the vehicle was there. That means knowing how much time passed between the frames being compared.

An error in either one carries straight into the speed. Here’s a simple example. Say a vehicle covers 60 feet between two frames, and the analyst believes that took exactly one second. That’s about 41 mph. If those frames were actually 1.07 seconds apart, the speed is closer to 38 mph. A difference of seven hundredths of a second changed the answer by about 3 mph.

Where the Timing Information Comes From

Surveillance video is stored on a recorder, usually called a DVR or an NVR. The difference mostly comes down to the type of cameras connected to it. For our purposes, both do the same job.

In the original video file, timing data lives in the metadata. Metadata is information stored inside the file that you don’t see on screen, such as when it was created and how the frames were spaced. Many surveillance systems also print a clock, called a timestamp, directly onto the picture.

Surveillance footage comes in many forms. Some systems print a timestamp directly on the picture, and the frame at the lower right is a phone recording of a playback screen.

What Happens When the Original File Is Gone

These recorders don’t keep footage for long. Most systems only retain the original recordings for 7 to 14 days before recording over them. If nobody exports the native file in that window, it’s gone.

What often survives is a phone video someone took of the playback screen. In our field this is called “video of video.” It keeps a representation of the images but loses the original file’s metadata, so the analyst can no longer look up when each frame was recorded. The printed timestamp may be the only timing information left.

That raises a fair question. How much can you trust that clock in the corner of the screen? That’s where the next post picks up.

Reference: Molnar, B., Terpstra, T., and Voitel, T., “Accuracy of Timestamps in Digital and Network Video Recorders,” SAE Technical Paper 2025-01-8690, 2025, doi:10.4271/2025-01-8690.