How do you record each egg to audit grading?

What does a per-egg record need to keep?

Each decision should leave three things: the class, a reason in plain language and the measurements that the reason uses. Without the measurements, the class is a claim that nobody can check later.

At Vigiovo, each egg comes out with a record in four parts: the class (clean, dirty, cracked or unverified), the scores for the three defect classes, the reasons and the measurements. The measurements include the dirt fraction, the crack length measured in reference radius, solidity and ellipse fill, the axis ratio, and whether the egg touches the image border.

The scores look like probabilities, but they come from a fixed formula applied to the measurements, not from an estimated probability. An audit should therefore look at the measurements and the reason, which show why the class was chosen.

How does an audit check one egg's decision?

The audit checks that the rule was applied correctly: it takes the recorded measurement, compares it with the limit and confirms that the class matches. Because the decision comes from fixed thresholds, anyone on the team can redo this step without depending on who ran the line.

The same image with the same configuration always gives the same result. If the image was kept, the audit can reprocess it and compare the output with the record. If the result changes, something in the configuration or the system has changed, and that change also needs to be recorded.

1. Check that the class matches the recorded reason and measurements.
2. Redo the decision for some eggs close to the limit, using the stored images.
3. Compare the reasons batch by batch to see whether the same defect always appears from the same source.
4. Note every configuration change with its date and reason, to explain differences between days and shifts.

Eggs close to a limit deserve special attention. The record keeps how far each measurement was from its limit, which separates clear cases from borderline ones, and a review should start with them.

What doesn't the per-egg record prove?

The record proves what the camera measured on the surface of the egg. It does not prove what is inside: the air cell, the yolk and hairline cracks are still out of reach of a surface camera, and those criteria depend on candling.

It also cannot say on its own which batch or shed an egg came from. That link has to exist in the company's own system. When egg types must have separate production flows, MAPA requires those flows to be described in the Autocontrole programme and in the product registration in PGA-SIGSIF (MAPA, question 27). A per-egg record is only useful for audit when it is tied to those flows.

What does MAPA require for verifying electronic equipment?

MAPA question 34 covers electronic equipment that gives results equivalent to traditional candling. It may be used if a manual check by a validated sampling system is planned, and the company must describe in its Autocontrole programme how it verifies the equipment, including the frequency, the way results are assessed and the way the verification is recorded.

A per-egg record helps with that verification, because it lets the team compare the equipment's decision with a manual assessment of a sample and review the cases where they differ. It does not replace the verification plan, which belongs to the company and must be written down. To compare the methods, see candling vs computer vision.

Whether and how this rule applies to a camera system that does not perform candling is a question for the Autocontrole programme and for the inspection authority, not something this text decides. Confirm it with your quality team before adopting any procedure.

Frequently asked questions

Is a per-egg record enough for an audit?

Not on its own. The record is strong evidence of each egg's decision, but an audit also depends on the Autocontrole programme, the described production flows and the link between the record and the batch. It is one part of that set, not the whole set.

Are the scores in the record probabilities?

No. At Vigiovo, the scores come from a fixed formula applied to the egg's measurements, not from a probability estimated from data. For an audit, check the measurements and the recorded reason, which show why the class was chosen.

Does MAPA require an individual record for each egg?

Question 34 covers recording the verification of electronic candling equipment, not an individual record for each egg. It asks that the Autocontrole programme describe the frequency, the assessment and the recording of that verification. Confirm the requirements that apply to your unit with your quality team.

Sources

1. Como o Vigiovo classifica ovos sem IA / How Vigiovo grades eggs without AI, Vigiovo
2. Perguntas e respostas sobre a Portaria MAPA nº 612/2022 (versão de 05/06/2023), Ministério da Agricultura e Pecuária (MAPA)