quality assessment of bologna sausages using machine vision
TRANSCRIPT
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Quality assessment of bologna sausagesusing machine vision
A. Mtrai, L. Baranyai**Corresponding: [email protected]
XIIth International Conference of Food Physicists6-8 July, 2016. - Debrecen, Hungary
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Bologna sausages
DefinitionDerived from Mortadella, made of homogeneous meat
pulp
with cheese, vegetables, etc.
Typical ingredientsPork (40%), beef (15%), fatty tissue (22.5%), water (22.5%)
Photo: Sauels AG, Kempen-Tnisberg, DE
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Food labeling
Regulation (EU) No 1169/2011 of the European Parliament and of the Council of 25 October 2011 (change 13 December 2016).
Article 18: List of ingredients
1. The list of ingredients shall be headed or preceded by a suitable heading which consists of or includes the word ingredients. It shall include all the ingredients of the food, in descending order of weight, as recorded at the time of their use in the manufacture of the food.
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Objectives
Development ofPortable and flexible vision system
Methods for evaluationRecognition (segmentation) of visible ingredients
Estimation of quantity in m/m%
EvaluationComparison of measured data with label information
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Vision System
Camera: CMOS 20.3 MP, Samsung NX300
Stative: Vanguard Alta CA 203AT tripod
Illumination: +2 12V / 50W filament lamp
BackgroundBlue paper sheet
Transparent plastic sheet
Exposure: auto
Resolution: 0.136 mm/px
Photo: Samsung
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Materials
5 type N = 20 (10 slices, both sides)
1., Parizer with cheese2., Turkey Parizer with cheese3., Parizer with cucumber4., Parizer with mushroom5., Meat bread (paprika Lyoner)
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Image processing
Color normalizationEnhance color differences
Uniform intensity
Segmentation:Dynamic k-means clustering
Linear type
Quadratic type
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Preprocessing
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Quantity estimation
Segmentation: grayscale clusters
Uniform slice thicknessv/v% estimation
Required: density
Density estimation from meat and fat:
PulpCheeseVegetable
Density (g/cm3)0.86 0.91.11.0
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Estimation results, m/m%
Sausage ...Cheese (p)Cheese (t)CucumberMushroomPaprika
Minimum9.514.214.3413.621.48
Median14.166.546.7319.262.05
Average13.786.496.5119.232.06
Maximum16.998.578.9025.322.95
Std. dev.2.781.230.993.280.42
Reference14712201.2
Sig.******
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Estimation results, m/m%
Sausage ...Cheese (p)Cheese (t)CucumberMushroomPaprika
Difference0.220.515.490.770.86
Sig.0.7300.0770.0000.3310.000
Sig.******
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Evaluation
Sausage ...Cheese (p)Cheese (t)CucumberMushroomPaprika
Difference0.220.515.490.770.86
CV0.2020.1900.1520.1710.202
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Suggestions
Better density estimationmeasurement
recipe
Portable box setup
Transillumination for transparent products
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Conclusions
Normalized color transformation and cluster analysis were able to segment ingredients.
The error of estimation was in the range of 0.22 5.49 m/m % where the smallest error was observed for cheese and the largest for cucumber.
Coefficient of variation was similar for all products.
The proposed method needs further improvement in setup and density data.
Obtained results are promising. System is recommended for non-destructive quality assessment during production.
Andrs Mtrai, Lszl BaranyaiSzent Istvn University, Faculty of Food Science1118 Budapest, Villnyi str. 29-43. - http://etk.szie.hu
Thank you for your attention!