Candling or computer vision: which should you use for egg grading?

What is candling and what does it detect?

Candling examines an egg against a bright light that shines through the shell and shows the inside. It is the method that reveals the height and mobility of the air cell, the yolk position, blood spots and fine cracks, including hairline cracks.

In Brazil it is required by law: RIISPOA lists candling among the procedures egg establishments must carry out (art. 223) and requires facilities and equipment for candling and grading (art. 45). Class A interior requirements, such as an air cell of at most 6 mm and a yolk visible only as a shadow, are checked at this step (art. 225).

What does computer vision detect in egg grading?

A camera over the conveyor sees the surface of each egg in reflected light. With image processing it detects adhering dirt, visible cracks, leakers and abnormally shaped eggs, and counts each class in real time.

It doesn't see the inside: the air cell, blood spots in the yolk and hairline cracks are out of reach for an ordinary camera. In return it measures every egg with the same criteria, without fatigue, and records the reason for each decision.

Candling or computer vision: which detects each defect?

The table sums up what each method covers. The computer-vision column describes what Vigiovo does today.

How do you combine candling and a camera on the same line?

The order that pays off most is camera first:

1. The camera over the conveyor removes dirty, visibly cracked and misshapen eggs, and marks as unverified the ones it can't see whole.
2. Candling works on the remaining batch, focused on what only it can see: the interior, the air cell and hairline cracks.
3. The camera's records show the dirty and cracked rate per batch and per house, which helps find the cause upstream.

This split uses the best of each method and doesn't trade a legal requirement for a promise. For the criteria behind each defect, see the guides on cracked eggs and dirty eggs.

AI or deterministic computer vision?

Most camera inspection systems use neural networks trained on thousands of images. Vigiovo uses deterministic computer vision: classical operations such as colour segmentation, watershed and morphology, with readable limits.

In practice the same image always gives the same result, every decision comes with the measurements behind it, and the system runs on an ordinary CPU at about 90 frames per second. The limit is that it only recognises the defects its rules describe. The full comparison is on the home page.

Frequently asked questions

Does computer vision replace candling?

No. Candling is mandatory in registered Brazilian egg establishments (RIISPOA, art. 223) and sees the inside of the egg, which an ordinary camera cannot. Computer vision complements candling by removing dirty and visibly cracked eggs first.

Does candling detect dirty eggs?

Poorly. Candling lights the egg from behind to see the inside, and light dirt on the shell barely shows that way. Surface inspection, by a person or a camera with reflected light, is what catches dirt.

What is the advantage of a camera over human visual inspection?

A camera applies the same criteria to every egg for the whole shift and records the class and reason for each decision. Human inspection varies with fatigue and volume and usually leaves no per-egg record.

Sources

1. Decreto nº 9.013/2017 (RIISPOA), Presidência da República (Brasil)
2. Delegated Regulation (EU) 2023/2465, marketing standards for eggs, EUR-Lex
3. Watershed segmentation, scikit-image
4. Como o Vigiovo classifica ovos sem IA / How Vigiovo grades eggs without AI, Vigiovo