RGB-D Depth and Colour Measure a Pallet in One 6 s Shot
Pallet volume data feeds billing, load planning and warehouse capacity planning. Measuring with a tape is slow and inconsistent, and the real problem is not only speed: data entry lags the work rhythm, the criterion varies between operators, and settlements have no defensible basis. This case uses an RGB-D 3D dimensioning camera that captures depth and RGB in one static exposure, fuses the two for volume measurement and presence detection, supports parcel classification and thin-item measurement, and calibrates with one key. The cycle is about 6 seconds, subject to site acceptance results.
Volume data has to be system-ready, not just accurate
Volume must be bound to a cargo ID and output as structured data before sorting and billing systems can consume it. The first objective is therefore usable, traceable data rather than a beautiful measurement. A number that arrives too late or uses an unclear criterion cannot be used for settlement, however precise it looks.
Why a static exposure beats a scanning gantry
Pallet cargo only needs to pause briefly in the measuring zone. One exposure captures the whole field, which suits cargo passing quickly on a conveyor and places low demands on conveyor start-stop behaviour; no motion axis has to be engineered for measurement. For large pallets, camera height and field are designed around the largest unit, covering the whole pallet in one shot.
Depth and colour cover for each other
The depth map provides the 3D geometry of the cargo surface, and length, width and height are computed from it. The RGB image provides texture and colour for region segmentation, parcel classification and edge recognition. Fused, geometry and appearance corroborate each other: when one channel fails on an unusual item, the other still supports a result. Capturing both in one exposure also removes synchronization calibration between separate devices.
Thin items are where depth noise shows
A thin item carries a weak signal in the height dimension and is easily buried in depth noise. The system refines the depth map, extracts the thin item contour and confirms its boundary from the height jump at the edge. This class of cargo is a common hard case at a dimensioning station and should be validated separately during selection.
Isolating the target pallet from its neighbours
Adjacent cargo and background usually sit beside the measuring zone. The system segments pallet, cargo and floor from the fused data and computes volume only for the target pallet. Floor undulation is absorbed by calibration parameters, and cargo height is top-surface height minus the reference height. One-key calibration fixes camera intrinsics, extrinsics and the height reference in a single operation, so ordinary maintenance staff can complete it.
Where the method needs validation first
Depth imaging struggles with transparent and strongly reflective surfaces. Where such cargo dominates, run a suitability trial before committing. High-volume conveyor lines can place measuring points at several entry points in parallel, so throughput scales with demand instead of pushing a single point's cycle.
Questions worth settling before deployment
What do depth and RGB each do here?
Depth computes the three dimensions; RGB handles segmentation and classification. Together, geometry is measured and appearance is sorted out.
Is a 6 second cycle enough for large items?
Static dimensioning does not scan, so one exposure captures the full field and cycle is largely independent of cargo size. Conveyor start-stop and compute time matter more, subject to site acceptance results.
What does one-key calibration mean?
Calibration is automated and simplified so the mapping between image and physical space is established without laborious manual steps, lowering the deployment barrier for site staff.
Where is this method unsuitable?
Where transparent or highly reflective cargo dominates, or where the zone cannot guarantee a brief stop, assess depth capability and work rhythm before adopting it.