Best practice: one object per scan

One object per scan is the most useful habit you can build. It applies whether you are sorting e-scrap, checking jewellery or looking at ore samples, and it is worth understanding why rather than just following it.

Note what the rule is and is not. One object per scan, several photos of that object. The guided scan takes multiple images, so showing a piece from the front, the back and close up is exactly right. What causes trouble is putting two different things in front of the camera and asking about both at once.

Why mixing objects goes wrong

When several distinct items share a frame, their features get read together, and a few specific failures follow.

Merged characteristics. A dull grey chip next to a gold-plated connector produces a mixed reading. The gold-coloured reflection belongs to one of them, but nothing in the image says which, so colour and lustre evidence gets attributed to the wrong object.

Focus and exposure go to one item. Your camera picks a focal plane and a brightness. Whatever it did not pick comes out soft or badly lit — so at least one object in a multi-object photo is always the worse-photographed one.

No way to know what you meant. Nothing in the photo says which item you care about. The larger or more central one tends to dominate, which means you can get a confident answer about the wrong object. A gold-flecked rock beside a plain one is the classic case.

Averaged results. If one item contains precious metal and another does not, you can get an answer in between that describes neither. You lose the clear signal a single-object scan would have given you.

The net effect is both kinds of error: missing metal that is there, and reading metal onto something that does not contain it.

Doing it properly

Isolate the object. Move everything else out of frame. Testing a handful of components? Lay out one, scan it, then the next. Each scan gets one subject.

Use a plain background. White paper or a clean, non-reflective surface. A busy background competes for attention and blurs the object's edges.

Fill the frame. Get close enough that the item occupies most of the image, while staying in focus.

One object means one object. Scanning a ring? Only that ring should be visible — not a second ring at the edge of the shot. Scanning ore? One specimen, not a handful.

Then add the angles. Once the object is alone in frame, use the multi-photo flow properly: the whole piece, the reverse or underside, and a sharp close-up of any marking. Angles that reveal something new are worth far more than several near-identical shots.

Cropping after the fact

Taken the photo and found clutter in the corner? Crop it.

• Start from a high-resolution original so cropping leaves detail rather than pixels
• Trim to frame the one object tightly, cutting out other items, empty space and fingers
• Do not cut off part of the object, or the specific area you want looked at
• Aim for the item nearly filling the cropped frame

No cropping tool handy? Slide a sheet of paper behind the item before you shoot. Same effect, less editing.

Groups of genuinely identical items

Sometimes you want a read on a batch — a pile of gold-plated pins, say, where the question is about the lot rather than any one pin. It can be done, with care:

• Keep them uniform. Same type, same apparent material. Mixing a pin, a brass hinge and a silver contact reintroduces every problem above.
• Spread them out. A flat, non-overlapping layout. A heap reads as one confusing shape.
• Expect an average. One odd item among the rest skews the whole result.
• Scan individually when it matters. A group photo is a quick first pass. Anything you need to rely on gets its own scan.

Even among identical pieces, one tarnished item shifts the colour profile the camera records. Isolating one at a time remains the more reliable approach.

The summary

• One object per scan, several photos of that object
• Plain background, object filling the frame
• Sharp, well-lit, high resolution, every marking readable
• Groups only when the items are genuinely the same, and never when the answer matters

A clean single-object scan gives the analysis the clearest version of your question, which is the best way to get a useful answer. It is still an estimate — a photo cannot establish purity — but a well-posed question is where a useful estimate starts.