Developer Terms and Conditions General Discussions Projects and Ideas Meat Identification / mix quantification.

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
  • #1000

    I would love to be able to scan minced meat and be able to see what animal species are there, and to what % and how much fat  %, but I expect that the difficulties will be many.
    Joints of meat may be easier, but results may depend on the cut and age of the sample.
    Minced meat would be more topical, especially after the horse meat in burgers scandals.
    I see the hurdles as being:-
    a) the mix of fat and lean in any scan, with mince coarseness compared to the scan area being critical.
    b) irregular mix ratio within any batch.
    c) is there enough difference between pure samples anyway?

    I guess that we should start with c) ; if that looks promising but a) and b) are an issue, then perhaps we would need to grind and mix the sample before scanning.
    At least reference samples should be easy – if we find butchers that we can trust  :unsure:


    I’m also interested in meat identification. I agree that we should start with pure samples first. My team can collect some samples in USA and Asia. Keep me in the loop.


    Hi Khoa, Great to see a common interest.
    I suggest that we make a few independent samples first to get a feel for the data.
    After that we could share learnings and agree an attribute model and scanning protocol.
    I already see a big discrimination between spectra from lean meat and pure fat, that makes me think that fat % should work, depending on sample. See attached file.
    I will make another visit to the butcher soon to get some different “known” meats.
    I don’t expect this to be an easy project, as meat is coarsely granular and changes a lot with age and storage conditions.

    mfg Roger

    You must be logged in to view attached files.

    Awesome!!! CP emailed me  and said that my SCiO was on the way. Will let you know when I have some initial scans.

    Yes, we should get reliable and independent data first on meat. If possible, I suggest we label them as much details as we can:

    1. species and geographic origin (
    2. cuts (
    3. freshness (fresh meat or frozen)

    After that, the analyzing steps can follow in various directions such as quantification of fat, classification of meat species, meat brands and possibly geographic origin (Kobe beef vs Australian beef).




    Hi Khoa,
    I like your model ideas, we may need to include a couple of fields for how much marbling the sample has (or limit scans to lean areas only), freshness (colour?)
    I have tested ONE sample each from Beef, Pork and Lamb. There is considerable overlap, so lots more samples are required – and I am on a weight loss diet! This will not be easy.
    Fat % in minced meat looks relatively easy, but I need to see how much sample prep is needed.

Viewing 5 posts - 1 through 5 (of 5 total)
  • You must be logged in to reply to this topic.