ANNEX IV U.K. Authorisation of automated grading methods for beef and sheep carcasses referred to in Article 10

PART A U.K. Conditions and minimum requirements for authorisation

1. The Member State concerned shall organise an authorisation test for a jury composed of at least five licensed experts in classification of carcasses of bovine or sheep animals. Two members of the jury shall come from the Member State performing the test. The other members of the jury shall each come from another Member State. The jury shall comprise an uneven number of experts. The Commission services and other Member States` expert may attend the authorisation test as observers. U.K.

The members of the jury shall work in an independent and anonymous way.

The Member State concerned shall nominate a coordinator of the authorisation test who shall:

(a)

not be part of the jury;

(b)

have satisfactory technical knowledge and be fully independent;

(c)

monitor the independent and anonymous functioning of the members of the jury;

(d)

collect the classification results of the members of the jury and those obtained by using the automated grading methods;

(e)

ensure that, during the entire duration of the authorisation test, the classification results obtained by using the automated grading methods shall not be available to any of the members of the jury and vice versa nor to any other interested party;

(f)

validate the classifications for each carcass and may decide, for objective reasons to be specified, to reject carcasses from the sample to be used for the analysis.

2. For the authorisation test: U.K.

(a)

each of the classes of conformation and of fat cover shall be subdivided into three subclasses;

(b)

a sample of at minimum 600 validated carcasses shall be required;

(c)

the percentage of failures shall be no more than 5 % of the carcasses that are fit for classification by using automated grading methods.

3. For each validated carcass, the median of the results of the members of the jury shall be considered as the correct grade of that carcass. U.K.

To estimate the performance of the automated grading method, the results of the automated grading method shall, for each validated carcass, be compared to the median of the results of the jury. The resulting accuracy of the grading by automated grading methods is established by using a system of points that are attributed as follows:

	Conformation	Fat cover
No error	10	10
Error of one unit (i.e. one subclass up or down)	6	9
Error of two units (i.e. two subclasses up or down)	– 9	0
Error of three units (i.e. three subclasses up or down)	– 27	– 13
Error of more than three units (i.e. more than three subclasses up or down)	– 48	– 30

With a view to authorisation, the automated grading methods should achieve at least 60 % of the maximum number of points for both conformation and fat cover.

In addition, the classification by using the automated grading methods must be within the following limits:

	Conformation	Fat cover
Bias	± 0,30	± 0,60
Slope of the regression line	1 ± 0,15	1 ± 0,30

Where during an authorisation test more than one carcass presentation is used, the differences between those carcass presentations shall not lead to differences in the classification results.