hul 2004 
matische 
rfassung. 
(PFG), 
yf semi- 
Third 
an-Made 
tzerland, 
ma, The 
)perative 
er Vision 
pke, C., 
ta Using 
es of the 
ormation 
lygonal 
ics and 
f spatial 
rence on 
berCity- 
ery and 
' Photo- 
sciences, 
y Profile 
of Man- 
> Federal 
mo GIS 
VISION-BASED SYSTEM FOR QUALITY CONTROL OF SOME FOOD PRODUCTS 
O. Alhusain *" *, Z, Tóth?, Á. Rakusz *, L. Almâsi *, B. Farkas © 
“ University of Technology and Economics, Department of Photogrammetry and Geoinformatics, H- 1521 
Muegyetem rkp. 3., Budapest, Hungary - alhusain@eik.bme.hu, tothzoli@index.hu, aaaah@mailbox.hu 
? HAS-TUB Geoinformation Research Group, H- 1521 Muegyetem rkp. 3., Budapest, Hungary 
* Miklós Zrínyi National Defence University, H- 1581, pf. 15, Budapest, Hungary - farkasb@zmne.hu 
Commission I1I, WG III/5 
KEY WORDS: Industry, Inspection, Pattern, Recognition, Vision 
ABSTRACT: 
Mass production of crop, produce, and food staff has resulted in great increase in the efficiency of food producing plants. This in 
turn has led to considerable decline in food prices. However, the mass production of food was also associated with two major 
problems. The first one is the decline in food quality, and the second one is the *waste" problem associated with processing and 
preparation operations. The wastage in many cases is a direct consequence of the quality problem, where the quality decline reaches 
unaccepted limits. Hence, is the need for quality inspection and assurance mechanisms to be installed in the production lines of such 
mass food processing and producing plants. 
In this paper, quality control and inspection system relying on image processing techniques will be discussed and presented in 
details. 2D and 3D visual characteristics are collected about three kinds of food products that are in one way or another go through a 
mass production process. These products are Arabic style pita bread and Mexican tortillas. Visual characteristics of interest for this 
study are collected during the baking phase for the first two products and during classifying and sorting process for the last one. 
Characters measured for the products were size (width, length, volume, and area), shape, and color (dominant color, localized colors, 
average color). 
Imaging system utilizing 5-megapixel digital camera was used to acquire the images for this study. Different image processing 
procedures like filtering, binarizing, zoning, masking, and enhancement were used to derive the quality control parameters from the 
visual characters of the products under investigation. This study has shown that machine-based inspection of food products can be 
implemented effectively, reducing or even eliminating the need for both intensive human intervention and addition of conditioning 
chemicals to assure quality. The concept and results presented in this study can be applied in solving more complicated pattern 
recognition problems. 4 
1. INTRODUCTION : 1. Sensor system: Which is generally CCD or CMOS 
camera, the resolution in these cameras vary from one 
system to the other, common resolutions found in 
these systems range from sub mega pixel on the low 
side to almost 10-20 mega pixels on the high side. 
Recent developments in machine vision and supporting 
technologies has resulted in general acceptance of the feasibility 
and profitability of implementing visual inspecting systems in 
quality assurance operations of food producing lines. Machine 
vision benefited the most from the increase in processing and 
storage powers of modern chips, and from the emergence of 
megapixel sensing and imaging devices. Machine vision 
technology utilizes image processing techniques for the purpose 
of extracting visual features about an object for a variety of 
qualitative, quantitative and control applications. 
During the last few years, the trend in designing and building 
machine-based inspection systems has almost always followed 
one of two patterns. These are the traditional sensor (camera) — 
computer — interface pattern, and a newer pattern- the so called 
smart camera pattern. The two patterns are widely presented 
and deeply discussed in recent literature ( ICD, 2003; Wilson, 
2003). 
1.1 Sensor (camera) — computer — interface pattern 
The inspection system consists of four distinct entities, these 
are: 
  
* Corresponding author. 
4T] 
Inspection data is fed to a computer system for 
processing, evaluation and decision making. 
The computing facility: In most cases it comprises a 
complete computer system with its operation system, 
application programs and/or task-specific programs 
(library). The hardware is built around Pentium 
processors or power PC ones. However, it seems to be 
that Pentium-dependant implementations are more 
common in scientific laboratories and academia 
mostly for flexibility reasons. While power PC 
implementations are more popular in real-life 
production lines mainly for stability characters of 
software implemented on these systems. 
Nevertheless, operating efforts and kills required to 
operate a power PC — based system are less 
demanding than those required to operate a similar 
Pentium — based system. 
Input/Output, and Networking facility: These 
facilities are easy to implement since they rely on 
utilizing the computer system’s facilities to achieve