3D ToF Cameras for Pallet Detection and Precise Docking

Give autonomous forklifts depth data for pallet pose estimation, fork-pocket localization and closed-loop approach.

Plan a Pallet Evaluation View Camera Options

Locate the pallet and fork pockets

Use depth geometry to identify pallet position and pocket regions within the camera view.

Estimate alignment before insertion

Provide spatial data for calculating pallet offset, orientation and approach corrections.

Support a closed-loop docking workflow

Feed updated perception data into the host system during alignment, approach and pickup validation.

Know where the camera ends and the application begins

DOMI supplies the depth-sensing layer. Your application or system integrator turns those observations into pallet decisions, vehicle actions and recovery logic.

DOMI camera provides

Depth data and integration building blocks

Depth map
IR image
Optional RGB image
Registered XYZ point cloud
Calibration parameters
SDK and host interfaces

Your application system implements

Perception, control and exception behavior

Pallet and fork-pocket detection
6-DoF pallet pose estimation
Camera-to-fork coordinate transform
Approach planning and vehicle control
Pickup confirmation and exception handling
Safety-rated functions and validation

Safety-rated functions must be implemented and validated within the vehicle's certified safety architecture.

From perception data to a controlled docking loop

The camera supports each handoff. The host system decides what the next action means and when the vehicle should retry or stop.

01

Detect Pallet

Find the pallet geometry in the current camera view.
02

Locate Pockets

Identify the left and right fork-pocket regions.
03

Estimate Pose

Calculate position and orientation relative to the camera.
04

Align Vehicle

Use updated offsets to correct the approach trajectory.
05

Insert Forks

Move through the insertion step under vehicle control.
06

Confirm or Retry

Validate pickup state or return to a defined retry path.

Reliable docking depends on perception quality, calibration, mounting geometry, control strategy and validation under the target operating conditions.

Validate against the pallets and conditions that matter

Pallet geometry, surface properties, occlusion, ambient light and mounting position can all affect the available depth data. Evaluation should reproduce the target operating environment.

RGB, depth and 3D point-cloud views of an autonomous forklift approaching pallet fork pockets — RGB, depth and point-cloud views for pallet and fork-pocket evaluation.

Skewed pallet

Check pose estimation when the pallet is not square to the vehicle.

Damaged or deformed pallet

Test how missing boards, bent runners and irregular edges affect detection.

Partial occlusion

Reproduce blocked pockets, loads and nearby objects from the operating aisle.

Shrink wrap or reflective surfaces

Measure valid depth and define a response for ambiguous returns.

Dark materials

Include low-return surfaces at the distances and angles that matter.

Changing ambient light

Validate the depth signal across the real indoor or semi-outdoor light range.

Optical and motion qualification

Design for the optical edge cases first

Pallet docking is a close-range, reflective and moving perception problem. Qualify the physical sensing envelope before tuning the application algorithm.

Fork-pocket cavities and multi-path interference

Pallet pockets are concave optical cavities. Reflections can distort depth at corners, so check valid-depth or confidence behavior, edge filtering and repeatability with representative pocket geometry. Product-specific anti-MPI behavior must be confirmed with DOMI.

850 nm or 940 nm is a system choice

Compare emitted power, sensor response, ambient infrared, target reflectivity and working range in the real scene. Do not transfer the published range of one wavelength configuration to another.

Near-field insertion and terminal approach

The listed options begin around 0.2-0.3 m. Characterize the final insertion distance and near-field behavior, and consider an independent proximity or limit signal for terminal stop logic.

Motion, frame rate and timing

Frame rate is not end-to-end control latency. Qualify vibration, exposure behavior, host timestamps and motion during the 10-30 fps path. Confirm hardware trigger or multi-camera sync for the selected module.

3D ToF, stereo vision or 3D LiDAR?

Choose from the scene geometry and result you need, not from the sensor label alone. The comparison below frames the engineering questions to bring into evaluation.

Technology comparison for pallet docking perception
Technology	Why teams consider it	Docking questions to qualify
3D ToF	Dense per-pixel depth for near-field geometry and low-texture surfaces.	Ambient infrared, reflectivity, multi-path effects and near-field range.
Stereo vision	Depth from correspondence when texture, baseline and lighting support matching.	Low-texture wood or plastic, occlusion, baseline and compute budget.
3D LiDAR	Longer-range spatial context with a different density, scan and integration tradeoff.	Pocket-edge coverage, close-range behavior, scan timing and vehicle latency.

Fit the camera to the vehicle, not the other way around

Working distance, optical coverage and coordinate frames shape the quality of the docking data before any algorithm runs.

Forklift mast and carriage with a forward-facing depth camera, fork-pocket field of view and camera-to-fork calibration concept — Mounting, field of view and camera-to-fork calibration concept for pallet docking.

Working distance

Define the minimum and maximum distance from the camera to the pallet and pockets.

Required field of view

Cover the pallet width and approach envelope without wasting pixels outside the task.

Pixel coverage at the pockets

Confirm that pocket edges and depth changes are represented at the target range.

Mounting and calibration

Compare mast-mounted, fork-carriage mounted and over-the-fork positions, then qualify the camera-to-fork coordinate transform.

Host interface and compute

Match outputs, bandwidth, SDK path and processing platform to the vehicle architecture.

Choose a camera path for evaluation and deployment

Start with one representative platform, then compare the interface and optics against the production envelope.

DOMI camera specification and evaluation comparison matrix
Evaluation role	Model	Wavelength	Published range	Interface	Output or integration focus
Algorithm POC and dataset	DM-RGBD5003A	940nm	0.2-5.0 m	USB 2.0	RGB, depth and registered point cloud. Check ROS / ROS 2 version requirements.
USB depth integration	DM5005A	850 nm or 940 nm	0.3-5 m at 850 nm; 0.2-2 m at 940 nm	USB 2.0 / UVC 1.2	Point Cloud, Depth and IR output
Indoor embedded OEM	DMOM2808D	850 nm	0.25-4.5 m	MIPI CSI-2, 2 lanes	320 x 240 module path. Confirm output mode and calibration flow.
Docking and semi-outdoor	DMOM2508CL	940 nm	0.2-2.0 m	MIPI	Near-range pocket tracking candidate. Validate light and material conditions.

Integration choice DM5005A USB depth and IR integration View details

Embedded OEM choice DMOM2808D 850 nm MIPI module path View details

Near-range docking choice DMOM2508CL 940 nm MIPI module path View details

DMOM2808D and DMOM2508CL are conditional MIPI choices. Compare range, wavelength, field of view and actual light and material conditions before selecting the production module.

Need another RGB-D configuration? Explore DM-RGBD5002A

Published product specifications are a starting point. Confirm strong-light behavior, software versions and project-specific performance during evaluation.

Turn an application question into evidence

A staged evaluation turns a camera specification into a sensing path your robot team can test and own.

01
Define the operating envelope

Pallet type such as Euro pallet (EUR 1), GMA, CHEP, block or stringer, plus distance, lighting, speed and mounting constraints.
02
Select the camera configuration

Match range, field of view, wavelength, output and host interface.
03
Capture representative data

Include nominal pallets and difficult conditions from the real site.
04
Validate the complete loop

Evaluate detection, calibration, alignment, insertion and exception recovery together.

Questions engineers ask before evaluation

Short answers to common camera, integration and operating-condition questions.

Does DOMI provide a complete pallet detection algorithm?

DOMI cameras provide depth, IR, optional RGB, point-cloud and calibration data. Pallet detection, fork-pocket localization, pose estimation and vehicle control are normally implemented in the customer or system-integrator application.

Should I start with an RGB-D camera or a depth-only module?

Start with RGB-D when you need fast data capture, appearance cues and an evaluation path through RGB-D or XYZ point clouds. Choose depth-only when the production design has a clear need for depth, IR, compact integration or a MIPI host path.

How should the camera be mounted on an autonomous forklift?

Mounting depends on the working distance, field of view, fork geometry, vibration, cover glass and calibration strategy. Use the recommended mounting concept as a starting point, then validate the final camera-to-fork transform on the vehicle.

What affects pallet-pocket localization performance?

Pallet geometry, distance, pixel coverage, surface properties, occlusion, ambient infrared, mounting position, calibration and host-system latency all affect the available depth data. Reproduce the target operating envelope during evaluation.

Can the system work with damaged pallets or shrink wrap?

It may be possible, but these conditions must be tested with representative pallets and materials. Define valid-depth thresholds, retry behavior and exception handling in the application rather than assuming every surface will return usable depth.

What data can the camera provide to the host system?

Depending on the model, the camera can provide depth, IR, optional RGB, registered RGB-D or XYZ point-cloud data, calibration parameters and a model-specific SDK or host interface. Confirm exact outputs in the product documentation.

How does the ToF camera handle multi-path interference inside pallet fork pockets?

Pallet fork pockets are concave optical cavities, so multi-path reflections can distort depth at corners. Evaluate the selected module's valid-depth or confidence behavior, edge filtering and repeatability with representative pockets. Product-specific anti-MPI behavior must be confirmed with DOMI.

Does DOMI provide ROS 2 drivers for AMR and AGV controllers?

DOMI product materials list ROS and ROS 2 development paths for representative RGB-D evaluation. Confirm the selected model, ROS 2 distribution, driver or node availability and message contract, including any Humble, Iron, Nav2 or MoveIt requirements, before integration.

How do I start an evaluation?

Share the pallet type, working distance, vehicle platform, mounting position, lighting, target conditions and preferred interface. DOMI can then help identify a representative camera configuration and a validation path.

Plan your pallet evaluation

Find the right sensing path for your pallet docking system

Share your pallet type, operating distance and vehicle integration requirements. Our team will help evaluate the camera configuration and validation path.

Camera and configuration fit for your working distance
Validation priorities for pallet, lighting and mounting conditions
A clear integration path from depth data to vehicle control

Tell us about your pallet docking application

A few details are enough to start the conversation.