A 3D camera can fit the whole part in its image yet still miss the surface you need to measure. Bin walls, projector shadows, changing point spacing and mounting clearance all affect whether a position works. This guide helps you choose a physical placement from the inspection volume, smallest feature and measurement tolerance. You will then validate that placement at the nearest, farthest and most difficult surfaces before committing to hardware.
1. Define the inspection volume before choosing a distance
Start with a volume, not a single working plane. Record the width, length and height occupied by every acceptable part position. Include tall stacks, tilted parts, fixture variation and any surface that must remain measurable as the container empties.
Keep these requirements separate:
- Coverage: every required surface lies inside the usable measurement volume.
- Sampling: the smallest feature receives enough valid depth samples for the inspection algorithm.
- Measurement error: measured dimensions or positions remain within your allocated error budget.
- Visibility: the relevant receiving optics and, where applicable, projector can see the surface.
- Clearance: the camera and mount remain outside the approved machine and tooling envelope.
Working distance may be specified from a housing datum rather than an optical center. Use the manufacturer's documented reference when checking operating limits. Do not compare distances measured from different origins.
The guide to depth image field of view explains why a visible image boundary does not necessarily represent usable depth coverage.
2. Estimate coverage at both ends of the volume
For a fixed perspective camera, a plane farther from the optical center generally spans a larger field of view. A pinhole estimate helps screen layouts when the field angles are known. It does not establish the camera's valid depth range.
Here, Z is axial distance in the calibrated optical coordinate system. The equations describe planes perpendicular to that axis. They do not account for projector overlap, occlusion, invalid border regions or surface-dependent depth failures.
Check coverage at the nearest surface first, because its geometric field is smaller. Then check the farthest surface for point spacing and measurement performance. A tilted inspection volume also needs intermediate cross-sections; its corners may occur at different depths.
Add explicit margins for placement variation and border exclusion. For example, a 400 mm wide envelope with a 25 mm allowance on each side requires 400 + 2 × 25 = 450 mm of usable coverage. That allowance is an application choice, not a universal recommendation.
3. Translate the smallest feature into a sampling requirement
A useful first estimate divides field width by the number of useful depth samples across that width. Use the depth output grid, not the color image dimensions.
Choose the desired sample count through algorithm trials. Five samples across a feature can be a starting screening target, but it is not a guarantee of detection or dimensional accuracy. Edge filtering, invalid points and correlated depth estimates can reduce the useful information.
The camera-resolution calculator below can help with this arithmetic. Treat its output as a sampling estimate, not a prediction of depth uncertainty.
Sampling and accuracy are different quantities. A dense cloud can contain biased depth, while a fitted surface can sometimes be located more precisely than the point spacing suggests. The deciding test is whether the actual algorithm measures representative parts correctly throughout the inspection volume.
4. Compare mounting positions by visibility, not appearance
Begin with a near-normal view when the task concerns upward-facing surfaces. This often simplifies coverage and coordinate interpretation. It is not automatically the best arrangement for every material or container.
| Position | Useful starting case | Main trade-off | What to verify |
|---|---|---|---|
| Near-normal overhead | Top surfaces and shallow trays | Deep walls can hide lower surfaces | Floor corners and surfaces beside walls |
| Offset with moderate tilt | Tool access or troublesome reflections | More occlusion and unequal range across the scene | Near rim, far wall and smallest features |
| Multiple fixed views | Opposing faces or persistent hidden regions | Calibration, timing and overlap become harder | Registration error and cross-camera interference |
| Movable acquisition position | Several distinct inspection views | Motion adds settling time and pose uncertainty | Measurement quality after each approved move |
For structured-light measurement, a point generally needs illumination from the projector and visibility to the receiving optics. Seeing a surface in the intensity image alone does not prove that it can produce depth.
Reflective bin walls can redirect projected light and corrupt measurements. An offset placement may reduce those paths, but it can also introduce shadows. Compare candidate positions using unchanged capture settings before attributing an improvement to geometry.
For a face tilted relative to the optical axis, point spacing along the sloping surface also increases. Avoid choosing an angle solely because the preview looks more complete.
5. Work through a placement example
Consider a hypothetical inspection envelope measuring 400 × 300 mm, with surfaces spanning axial distances from 600 to 900 mm. These are application assumptions, not specifications for any product shown below.
Allow 25 mm around each lateral edge:
Suppose a candidate optical geometry provides a 480 × 360 mm field at the near plane. It passes the geometric coverage screen, although the vertical allowance beyond the required envelope is limited.
Using the fixed-optics estimate:
Now assume a useful depth grid of 1600 × 1200 samples. At the far plane:
If preliminary algorithm trials call for five samples across that feature, the far plane misses the screening target. The corresponding grid requirement is:
Possible responses include a different camera, a smaller inspection region or separate acquisition positions. Moving the camera closer is useful only if the near-plane coverage and documented operating range still pass.
Finally, confirm the usable depth footprint experimentally. The calculated rectangle does not prove that the bottom corners receive valid measurements through the bin opening.
6. Set a measurement error budget separately
Do not assume that depth error scales directly with working distance for every sensing method. Range, surface reflectance, exposure, reconstruction settings and calibration can change performance differently across devices.
Separate repeatability from bias. Repeatability describes variation under repeated conditions; bias describes disagreement with a trusted reference. Neither alone establishes total application accuracy.
For an inspection tolerance of ±1.0 mm, you might allocate a tighter ±0.5 mm internal acceptance target to preserve integration margin. That is a deliberate project allocation, not a standard requirement.
A simple screening calculation is:
This example misses the internal target despite apparently modest frame-to-frame variation. The calculation is not a guaranteed error bound: it does not capture untested materials, correlated frames, reference uncertainty or systematic changes with pose.
Repeat the measurements after independent part repositioning, not just repeated captures of one static scene. For coordinate-based inspection, include the checks in camera extrinsic calibration accuracy.
7. Choose hardware only after the geometry screen
Use the 3D sensors category to identify the sensing approach, then narrow candidates with the camera finder. Confirm documented working range, usable depth coverage, acquisition behavior and measurement performance for your surfaces.
The MV-DPS200P-02 is a projected structured-light 3D camera. It is an evaluation candidate, not a confirmed match for the numerical example. Obtain the relevant operating data before assigning it to the proposed layout.
- Type
- Stereo / structured light
Choose mounting hardware around attachment geometry, supported load, adjustment range and stiffness. The camera stands category and accessories finder help organize that selection.
The LS-MNT-0040 is a slide-table camera stand. Confirm its payload capability, mounting interface and adjustment envelope before using it for a particular camera assembly.
- Type
- Camera stands
Include cable forces and adjustment locks in the mounting review. A movable slide is useful during trials, but the final position must remain stable after adjustment and normal machine activity.
8. Validate the nearest, farthest and hardest surfaces
Use a controlled test matrix rather than one attractive point cloud. Start with fixed reconstruction settings and record the camera pose, part pose, lighting state and software configuration.
- Test the empty or lowest scene. Inspect floor corners, recessed details and the farthest required measurement surfaces.
- Test the tallest acceptable scene. Confirm near-range validity, full coverage and clearance around the camera assembly.
- Place reference features across the field. Check the center, borders and corners at representative depths.
- Challenge visibility. Put parts beside walls, beneath lips and behind neighboring parts where production allows those arrangements.
- Challenge materials. Include dark, reflective and mixed-finish examples from the intended part population.
- Repeat after repositioning. Verify that apparent success does not depend on one favorable orientation.
Record valid-point coverage specifically inside each measurement region. A high valid-point percentage for the entire image can conceal failure on the feature that controls acceptance.
Use missing-point diagnosis when coverage fails locally. If a reference plane appears systematically tilted, investigate laser triangulation calibration checks before compensating by moving the mount.
After choosing a stable placement, tune acquisition settings. The guide to capture-time depth resolution and field of view addresses software-controlled changes separately from physical positioning.
9. Recognize when one camera position is the wrong answer
A single placement may cover the container opening but never see behind an overlapping part. More pixels cannot recover an occluded surface, and longer exposure cannot correct a blocked projector path.
Use separate views when required surfaces face opposing directions or remain hidden across normal part arrangements. If clouds must be combined, account for registration uncertainty and motion between captures. The guide to merging multiple depth frames explains that separate processing problem.
The closest possible position is also not always best. It can clip tall parts, leave insufficient adjustment margin or force critical features into unreliable border regions. Conversely, moving farther away to gain clearance can make small features inadequately sampled.
If no position passes coverage, sampling, measurement error and clearance together, change the inspection architecture rather than relaxing an essential acceptance criterion.
10. Complete the placement checklist before ordering
- Distance datum recorded: drawings and measurements use the same documented reference.
- Full volume checked: nearest, farthest and intermediate surfaces pass coverage tests.
- Smallest feature verified: sampling supports the actual inspection algorithm.
- Error target passed: bias and variation are measured against a suitable reference.
- Optical paths reviewed: receiving optics and projector paths remain clear where required.
- Mounting reviewed: attachment, stiffness, adjustment locks and cable forces are acceptable.
- Clearance approved: tooling and machine motion remain compatible with the installed assembly.
- Configuration saved: the validated pose and capture settings can be restored.
Use the stand configurator once the mounting envelope is established. If the geometry remains uncertain, send your inspection volume, smallest feature, tolerance and representative captures through application support before selecting the final hardware.
Questions
Where should 3D camera working distance be measured from?
Measure from the reference datum stated in the camera documentation. That datum may differ from the optical center used in projection calculations. Record the reference on your layout drawing, then check the nearest and farthest required surfaces against the documented operating range.
Does moving a 3D camera closer always improve accuracy?
A shorter distance can improve geometric sampling, but it can also reduce coverage or move surfaces outside the valid measurement range. Depth performance depends on the sensing method, surface and settings. Select the closest position that passes all application tests, rather than treating proximity as an accuracy guarantee.
What mounting angle should be used above a bin?
There is no universal mounting angle for bin inspection. Start with a position that covers the required volume, then compare wall reflections, projector shadows and tool clearance at alternative poses. Validate difficult corners and representative part arrangements rather than selecting an angle from the preview alone.
Is depth image resolution enough to choose a camera?
No. Image dimensions describe a grid, not the validity or accuracy of every depth estimate. Check useful depth samples across the smallest feature, then measure reference geometry at several depths and field positions. Include invalid points, edge behavior and repeated part placement in the assessment.
When should a second acquisition position be considered?
Consider another view when an essential surface remains hidden, or when no single position meets coverage and sampling requirements together. First establish that the failure is geometric rather than a settings problem. Additional views introduce timing, calibration and registration work that must enter the measurement budget.