Truck queue length at freight terminal gates
Sub-metre optical satellites can count stationary trucks queued at freight terminal gates, exposing throughput bottlenecks that port operators and logistics buyers rarely publish. Intra-day tasking turns snapshots into a queue-dynamics time series.
Sensors
- Planet SkySat: 0.50 m panchromatic, 0.72 m multispectral (RGB + NIR). Can be tasked multiple times per day over a fixed site; typical same-site revisit with full constellation tasking is 4 to 6 times per day. At 0.5 m, a standard 2.5 m wide truck cab is resolvable as a distinct object, which is essential for distinguishing a queued truck-and-trailer unit from a parked trailer without a cab.
- Airbus Pléiades Neo: 0.30 m native panchromatic, 1.20 m multispectral (6 bands including red-edge). The finest commercially available optical resolution at scale. At 0.3 m, cab rooftop markings and exhaust stacks are visible, reducing ambiguity between live trucks and abandoned trailers. Stereo tasking available; two satellites give up to twice-daily revisit over a target.
- Maxar WorldView Legion: 0.29 m panchromatic (WorldView-3 class). Six planned satellites increase revisit to roughly 15 passes per day at mid-latitudes over a given point, though not all passes yield cloud-free imagery. Archive depth from WorldView-2 and WorldView-3 predecessors extends to 2009, useful for multi-year baseline comparisons.
- Planet Basemaps (PlanetScope time series): 3 m multispectral daily mosaic. Individual trucks are not resolved at 3 m, but pixel-level change detection across a marshalling area can flag anomalous surface coverage consistent with unusual queue density, useful as a low-cost trigger for tasking a higher-resolution collect.
What a queue actually looks like from 500 km up
A standard European semi-trailer is roughly 2.4 m wide and 16.5 m long. At 0.5 m ground sample distance, that footprint covers about five pixels in width and thirty-three in length. The cab, which is typically 2.5 m wide and 2 m tall, casts a shadow of predictable geometry at known solar angles. This is the physical basis for automated truck detection: object-based image analysis segments candidate rectangles, then shadow length and cab-to-trailer ratio confirm whether a cab is present.
Without cab detection, a parked trailer without a tractor unit is indistinguishable from a queued truck in plan view. At 3 m resolution it is entirely invisible as a separate object. This is not a minor edge case: many terminals use drop-and-hook operations where trailers sit uncoupled for hours. Any queue-length methodology that does not explicitly handle this will overcount. Pléiades Neo at 0.3 m and SkySat at 0.5 m both provide enough detail for shadow-based cab detection; PlanetScope does not.
Revisit arithmetic and the intra-day queue problem
A single daily image tells you the queue length at one moment. Gate queues at busy container terminals can clear in under an hour or build to two-hour wait times within a single morning shift. A once-per-day collect from a standard medium-resolution constellation captures none of that dynamics.
SkySat's tasking model allows four to six collects per day over a fixed point when the full constellation is directed at a target. Pléiades Neo's two-satellite constellation can deliver two to three collects per day. WorldView Legion, when fully deployed, is designed for up to fifteen revisits per day at mid-latitudes, though cloud cover and off-nadir geometry reduce usable passes in practice. The honest position is that on a heavily clouded day at a tropical port, you may get zero usable optical images. SAR complements optical for all-weather detection, but at the resolutions currently available commercially, individual truck counting from SAR remains unreliable; it is a gap in the current toolkit.
For persistent monitoring, a practical architecture combines daily PlanetScope change detection as an alert layer with on-demand SkySat or Pléiades Neo tasking triggered when the change signal exceeds a threshold. This keeps tasking costs proportionate to events of interest.
The counting method and its honest limits
Published object-detection approaches for vehicle counting from very-high-resolution imagery typically use convolutional neural networks trained on labelled aerial or satellite imagery. Studies published in Remote Sensing (MDPI) have reported detection accuracies above 90 percent for trucks in open environments at sub-metre resolution, but accuracy degrades in several conditions: dense queues where trucks are bumper-to-bumper and shadows overlap; partially covered marshalling areas with canopies or trees; and scenes with high visual clutter from container stacks adjacent to the access road.
Queue length in vehicle count is the primary output. Converting that to wait-time estimates requires a ground-truth model of gate processing rate, which varies by terminal and shift pattern and is not observable from space alone. The satellite delivers the queue; the wait-time model requires the operator's or a third-party's gate-throughput data to calibrate. Buyers should be sceptical of any vendor claiming to deliver wait-time estimates without that calibration step.
Archive depth is a genuine advantage. WorldView-2 imagery exists from 2009, Pléiades from 2012. A multi-year retrospective analysis of queue patterns at a target terminal, correlated with published shipping schedules or vessel arrival data, can reveal structural bottlenecks that are invisible in any single observation.
Who buys this and what question they are actually asking
Port authorities use queue-length data to justify gate-hour extensions or additional lane infrastructure. The satellite record provides an independent audit of conditions that terminal operators may under-report to regulators or shippers.
Freight forwarders and beneficial cargo owners use it to route trucks away from congested gates, or to renegotiate dwell-time penalties with terminals when satellite evidence shows the delay was gate-side rather than carrier-side. Shipping lines use it to benchmark terminal performance across ports in a network, without relying on data supplied by the terminals themselves.
Commodity traders and supply-chain risk analysts use persistent queue anomalies as a leading indicator of port stress, particularly during labour disputes, infrastructure incidents or sudden demand surges. A queue that doubles over three consecutive days is a signal worth pricing.
Putting a number on coverage and latency
A single SkySat scene covers roughly 20 km², which is more than sufficient for any single terminal gate complex. Pléiades Neo strips can be ordered up to 400 km long at 0.3 m, covering an entire port precinct in one pass. Tasking-to-delivery latency for commercial very-high-resolution operators is typically two to six hours from collect to analyst-ready image, though priority tasking programmes can compress this.
Satellize integrates commercial tasking on client licence alongside open-constellation analytics. For this use case, the open-constellation role is limited to PlanetScope-class change detection as a trigger layer; the counting itself requires the commercial sub-metre collect. Our Tonga crop-estimation programme uses a structurally similar trigger-then-task architecture, applying open-data change signals to direct higher-resolution collection effort.
What the data cannot do, stated plainly
Cloud cover is the most obvious constraint. Tropical and high-latitude ports can see week-long overcast periods that block all optical collection. Night-time queues are not detectable with passive optical sensors; thermal infrared from commercial satellites at the relevant resolutions does not yet exist at the revisit rates needed.
The method counts trucks in a defined polygon. It does not identify the carrier, the cargo, or the booking reference. Linking a queue observation to a specific shipment requires integration with terminal operating system data or AIS vessel records, neither of which is delivered by the satellite layer alone. The satellite is an independent, tamper-proof counter. The interpretation of what caused the queue, and whose fault it was, requires additional data sources.
Typical figures
| Best available spatial resolution | 0.30 m panchromatic (Airbus Pléiades Neo); 0.29 m (Maxar WorldView-3 class) |
| Minimum resolvable target | Standard semi-trailer cab (approx. 2.4 m × 2.0 m) detectable with cab-shadow method at ≤0.5 m GSD |
| Intra-day revisit (tasked) | Up to 4–6 passes/day (SkySat); 2–3 passes/day (Pléiades Neo two-satellite pair); up to ~15/day (WorldView Legion at mid-latitude, cloud-permitting) |
| Spectral bands | Panchromatic + RGB + NIR standard; Pléiades Neo adds red-edge and coastal blue; all passive optical (no SAR for truck counting at this resolution) |
| Scene/strip coverage | SkySat: ~20 km² per scene; Pléiades Neo: strips to 400 km length at 0.3 m |
| Tasking-to-delivery latency | Typically 2–6 hours from collect to analyst-ready image for commercial VHR operators; priority programmes shorter |
| Archive depth | WorldView-2 from 2009; Pléiades from 2012; SkySat from 2013. Coverage of specific terminals varies by historical tasking demand. |
| Cloud and night limitations | Passive optical: zero usable data through cloud or at night. No commercial thermal-IR constellation at sub-metre resolution currently available. |
| Delivery formats | GeoTIFF (orthorectified); vector count layer (GeoJSON/Shapefile); time-stamped CSV queue count; API feed for integration with TMS or port community systems |
Analytics Satellize can run
| Point-in-time truck count per gate zone | Object-based image analysis with CNN-based vehicle detection; shadow-length filter for cab presence confirmation | GeoJSON layer with per-vehicle bounding boxes and cab-detected flag; summary count table per defined gate polygon |
| Intra-day queue-length time series | Multi-collect stack from same-day tasking; count differenced across collects to produce queue build and drain rates | Time-stamped CSV or dashboard chart showing queue length at each collect time; exportable to port community system |
| Queue anomaly alert | PlanetScope daily change-detection trigger (pixel-level surface change in marshalling polygon exceeds threshold); confirmed by on-demand VHR tasking | Email or API alert within agreed latency window; confirmation image with count overlay attached |
| Multi-terminal benchmarking report | Standardised counting methodology applied across a defined set of terminal gate polygons; normalised by gate-lane count where known | Monthly PDF or structured data report ranking terminals by median queue length and peak queue events |
| Historical queue-pattern baseline | Archive VHR imagery (WorldView-2/3, Pléiades) processed with consistent detection pipeline; aggregated by day-of-week, time-of-day, and season | Multi-year heatmap and summary statistics; suitable for infrastructure investment or contract negotiation evidence packs |
| Parked-trailer vs. queued-truck discrimination layer | Cab-presence detection via shadow geometry at ≤0.5 m GSD; cab-absent objects flagged separately and excluded from live-queue count | Annotated GeoTIFF with object class labels; separate counts for cab-present and cab-absent vehicle objects |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.