Skip to Main content Skip to Navigation
Conference papers

Real-time quality evaluation of pork hams by colour machine vision

Abstract : This paper presents the development of a real-time colour machine vision system which assesses the quality of raw hams by measuring meat colour and fat thickness. A specific machine vision device was set up and the algorithms use colour segmentation in the Hue, Saturation, Intensity (HSI) colour space. The detection of the muscles and fat in the images is carried out on the H and I components by using a thresholding method, on the S component by using a thresholding method followed by a connex component analysis and on the I component by a smoothing and a gradient analysis of the image, and an automatic thresholding method. Fat thickness is measured with regard to circular plate calibration. Muscle colour is measured in the HSI colour space by the use of 6 colour standards : Japanese Pork Colour Standards (JPCS). The method processes two kinds of ham (french and danish cuts) and runs in less than 3.5s.The results showed proper correlation between automatic and human measures of fat thickness (linear regressions on a set of 100 hams yielded R2 = 0.83 for the two kinds of measures) and proved the efficiency of image processing in grading hams according to their colour in real-time on production lines.
Mots-clés : CEMAGREF TERE GEAPA IRISA INRIA IAA
Document type :
Conference papers
Complete list of metadata

https://hal.inrae.fr/hal-02578973
Contributor : Migration Irstea Publications <>
Submitted on : Thursday, May 14, 2020 - 7:18:29 PM
Last modification on : Monday, August 31, 2020 - 3:06:02 PM

Identifiers

  • HAL Id : hal-02578973, version 1
  • IRSTEA : PUB00007973

Citation

D. Legeard, P. Marty-Mahé, J. Camillerapp, P. Marchal, C. Leredde. Real-time quality evaluation of pork hams by colour machine vision. Conference on Machine Vision Applications in Industrial Inspection VII, San Jose, USA, 25-26 January 1999, 1999, Seattle, United States. pp.138-149. ⟨hal-02578973⟩

Share

Metrics

Record views

14