How to estimate truck queue length from satellite imagery

If you run a border crossing or a terminal gate with no loop detectors, no license-plate reader, and no one physically walking the line with a clicker, you've probably already tried to eyeball queue length off whatever imagery you can get your hands on. It works, sort of, until the resolution isn't there or the pass was timed wrong and you're staring at a blur of roofs and shadows guessing where the line stops.

Here's the method planners use in practice, and where it falls apart without the right inputs.

What you're counting, and why pixel size decides everything

A queue length estimate from overhead imagery is really just vehicle counting plus known spacing. You identify the first truck at the gate, follow the line of vehicles back along the approach lane, and count units. Multiply by an assumed length-plus-gap per truck (commercial trucks plus following distance commonly land somewhere around 20-25 m, though this varies by corridor and truck mix) and you get a rough backup length in meters or kilometers.

The whole method depends on being able to tell one truck from the next. At coarse resolution (several meters per pixel) a line of trucks reads as a single gray smear along the road. You can see that something is backed up, but you can't count vehicles or tell a loaded flatbed from two cars nose to tail. For actual counting you need very high resolution imagery, roughly half a meter per pixel or better. At that scale a trailer is a distinct rectangle, cab shadows separate one unit from the next, and you can trace the line past bends and overpasses without losing it.

Timing the image matters as much as the resolution

A satellite queue length estimate is only as good as the moment the shutter opened. Queues build and clear over hours, sometimes faster during shift changes or when a crossing opens a second lane. An image pulled from an archive at the wrong hour gives you a clean, accurate picture of a queue that no longer exists by the time you're reading the briefing. For planning purposes, what you usually want is a pass tasked for the window you need, like the morning peak or right after a scheduled closure, not whatever the satellite happened to be overhead last Tuesday.

This is the part manual analysis struggles with most. You can request archive imagery and get lucky, or you can task a fresh pass and know the backup you're counting is the one happening now.

Doing the count yourself

If you're working this by hand: pull the sharpest image you can get for the time window that matters, zoom to the approach lane, mark the gate position, and walk the line truck by truck, flagging gaps where the queue breaks (a gap usually means the line hasn't backed up past that point, not that vehicles vanished). Measure the traced distance against the image's scale bar or known ground features, like a fence line or marked lane width, to convert vehicle count into a distance figure. Cross-check against any fixed reference, a bridge span, a known gate-to-rest-stop distance, anything with a documented length, to sanity-check your meters-per-pixel assumption before you report a number.

It's workable for a one-off check. It gets tedious fast if you need it daily, or for more than one crossing, or during a disruption when three people are asking you for the same number in different formats.

That's the gap Queue Length Monitor is built to close: a single VHR pass read into a numbered briefing note with the location, the queue length estimate, and the marked-up image the count was drawn from, so you're not re-deriving the method from scratch every time a crossing backs up. If a gate on your list has no sensor and you need a read on how far the line stretches back, take a look at what a single on-demand pass can tell you.