A tray is loaded with small components and the software returns a screen of green results. Before accepting the batch, we ask a separate question: did every loaded piece actually receive a complete inspection? With an Instant Vision Measuring Machine, touching silhouettes, overlapping parts and unrecognised objects can challenge the relationship between what is on the stage and what appears in the report. Good dimensional repeatability does not answer that counting question.

Our demonstration shows a batch measurement application. It does not certify automatic detection of every overlap or missing part in every software configuration. The efficiency comparison on the cover depends on the selected measurement task and comparison conditions. We qualify the complete inspection workflow using the customer’s components, including deliberately difficult loading arrangements.

Count three things, not just the green labels

We distinguish pieces loaded, objects recognised and pieces with all required results completed. These counts may differ. An object can be detected but fail one measurement; a component can be missed before measurement begins. A report containing only successful measurements is therefore insufficient evidence that everything placed on the stage passed. The expected count needs an independent, documented basis.

For a Desktop Instant Vision Measuring Machine, that basis might be a defined tray capacity with occupied positions checked, or a controlled count recorded during loading. We do not assume that a tray is full merely because it was full on the previous run. Empty positions, foreign material and a duplicated physical piece can all make a simple total misleading.

Touching and overlapping create different problems

Two separated parts normally have distinct visible boundaries. When they touch, their silhouettes may merge and the expected contour may no longer be available. An overlapping part can hide a boundary or an entire component. A top view cannot recover an edge that another opaque part physically conceals. More image processing does not remove that visibility limit.

We establish a practical separation rule through trials rather than prescribing one gap for all parts. Required spacing depends on the geometry, lighting, search regions and recognition method. Long projections or narrow slots may need more care than simple discs. Parts also need to remain separated while the stage is moved or the operator removes neighbouring pieces.

A Splicing Instant Vision Measuring Machine can extend inspection coverage for appropriate tasks, but stitching does not make overlapping material transparent. We retain the same requirement for each necessary feature to be visible and correctly assigned. A larger captured image is useful only when its contents can be interpreted reliably.

Challenge the routine with an incomplete batch

We test several plausible mistakes using controlled samples: one expected piece omitted, two pieces touching, a partly hidden piece and a part placed outside its permitted region. Where relevant, we also include a wrong but similar component. The purpose is to observe what the actual system reports, not to assume that an advertised recognition function covers every condition.

For each challenge, we record whether the programme stops, flags an incomplete result, rejects an object or silently excludes it. A silent exclusion deserves attention even if every remaining dimension passes. If the software cannot provide a dependable automatic check, an explicit operator verification or a physical loading aid must close the gap. We describe that dependency honestly in the operating method.

A Automatic Video Measuring Machine may suit a controlled fixture and programmed acquisition sequence. That arrangement still needs checks for empty positions and incomplete measurements. A fixed fixture reduces some loading variation; it does not prove that the right piece is present or that every required feature was acquired.

Keep no-result different from out-of-tolerance

A missing edge, failed recognition or interrupted sequence should not be converted into a zero, a blank treated as acceptable, or a value retained from the previous run. We review how the software and exported report represent these states. Their exact implementation varies, so the acceptance workflow should be demonstrated with the proposed configuration rather than inferred from a sample screenshot.

After correcting a loading problem, the operator needs a clear retry rule. We identify which pieces were remeasured, whether earlier values were replaced and how the final accepted batch is assembled. Mixing successful results from several partial runs without piece correspondence can create an apparently complete report that does not describe the actual outgoing set.

Additional views need the same piece identity

A Vertical and Horizontal Integrated Instant Vision Measuring Machine can support additional visible features where its optical arrangement is suitable. We still verify that results from different views belong to the same physical component. Two acceptable views of two different pieces are not a complete inspection of either piece.

Useful challenge trials also include normal production variation: surface finish, allowable rotation, ordinary contamination controls and the range of operators. We keep representative images of accepted loading arrangements and examples that require correction. Those examples make a spacing instruction easier to apply than a vague request to spread the parts out.

The trial output for a Vision Measuring System should include a reconciliation method, defined incomplete-result handling and a demonstrated recovery sequence. Send the part drawing, typical batch arrangement and current reporting requirements to handing3d@163.com. We at DONGGUAN CITY HANDING OPTICAL INSTRUMENT CO., LTD. can discuss a workflow in which a completed batch means every required piece has a traceable result.