This project is dedicated to development of diagnosis system of the metal
surface damage (corrosion, process defects, oil stain, topographical
irregularities, etc) or deformation of steel products and compatibility
solutions to enable large step improvements of productivity due to protect
surface quality claim in advance and save production loss by automatic surface
quality control system. The basic technical objective of this project phase I is
identification and explanation of major compatibility issues between surface
condition of any kind of metal and it’s digital image analysis by morphological
interpretation of color, shape, and texture of surface damage in real time.
Phase I is to demonstrate, by way of simulation tests of image analysis for
surface defects, the tradeoffs associated with conflicting requirements of
surface condition, for surface deformation from high stress or temperature of
production process, surface defect from corrosion inhibition and materials
fatigue life, wear resistance and corrosion resistance with lubricant for
process of milling, forging, pressing, etc. The anticipated results of the
project comprise key components of an advanced technology of diagnosis system of
surface damage analysis and methodology of study of surface defect with
mechanical properties. The project will result with: image capture and database
establishment procedures for surface damages, data expert system with data query
management, methodology of digital image analysis of morphological defects,
classification methods of defined images, automatic on-line prognosis system for
surface damage. The methodologies and experimental data that are conceived for
the proposed investigation will allow selection of optimal condition of surface
quality control and shape deformation.
This system can apply all kind of industrial production process and quality
control process for detection and prevention of surface damage or outside defect
of products.It will prevent the claim of production and increase the
productivity of company, which are producing final stage of metal relating to
surface.The superiority of this system is that 3 main digital image elements (color,
texture, and shape) can combine for image detection case by case and define by
multi-dimensional classification method.And it can adapt easily to user's
hardware with low cost of image capture system. Handy scanner in on-line system
will be useful for detection of surface damage and automatically can save all
data to user's main computer that are interpreting database. Especially for
detection of micro size defect, atomic force microscope or any kind of micro
detect device can be applied in on-line system.In case of internal defect
detection also, Digital signal from Magnetic Detector or Acoustic Device can be
used for defect analysis by non-destructive method.
Digital image analysis, Image capturing system, Export system, Morphology,
Surface damage, On-line Diagnosis, Multi-dimension classification, Lubrication,
Identification and significance of the problem or opportunity
The methods of evaluating surface damage and shape deformation existing at
present are based primarily on gravimetric, probing and visual assessments.
Their application often fails to identify adequately both the type and the
extent of damage. Moreover, most of these methods are targeted at laboratory
applications, frequently requiring specific preparation of specimens and
excluding all chances of in situ studies.
The review of publications indicates that the problem of evaluating the current
condition of a unit, the distribution and the area of damaged spots, the depth
of damage and the type of evolving defect, is specifically challenging for the
purposes of industrial monitoring. Evaluation of these characteristics allows to
determine the current condition of an article quite accurately and to conclude
which actions are suitable for prevention or reconditioning. These methods
should satisfy basic requirements of being compact, simple and implemental with
the available equipment, and they should yield easily interpretable data. Still,
though the problem is apparently essential, industries basically lack such
instruments, and expert visual examination remains the most frequent tool of
estimating surface damage. Hence, relevant means are needed to implement
effective algorithms, to process, analyze and store the necessary information
through the use of the state-of-the-art soft and hardware.
The progress of the hardware and software during the last decade allows us to
achieve a new solution of the problem of visual control of surface damage.
Incorporation of imaging devices into interfaces, means of producing standard
video signals, new compact digital solid body cameras and single-chip decoders
of video signals are primary tools. They are assisted by the advent and
intensive progress of multimedia capabilities in modern PC operating systems
supporting the input functions and the routing of the flow of video signals in
real time. Finally there are new effective algorithms of processing and
The morphological analysis of surface damage is one of the most widespread
diagnostic methods in surface science. It is known, that morphology of damaged
particles can indicate a type of damage process and current condition of
materials. The most of the methods based on the analysis of the defect images,
obtained by the optical or electron microscopy.
The shape of damaged particles is usually described in terms of its area,
perimeter, sizes of bounding rectangle, length of maximum and minimum radiuses
from the center to border of a particle and etc. [1,2,3,]. These features are
invariant to the rotation and translation of a pattern in a field of view and
have reasonable simple interpretation. However, they not always describe objects
in a unique fashion. It is possible to show there is a set of defect with
identical areas, perimeters and form factors but different under the shape. By
use of such type features the selection of representative set of them to assess
the defected particle shape is complicated by insufficient a prori knowledge
about correlation between shape and damage process. Simple adding new type of
features will not help to solve this problem. It was shown that there is no
algorithm for design and selection an optimal set of the features .
At last time, a considerable amount of interest has been devoted to
characterizing shape of contours by Fourier descriptors [5,6,7]. In this
approach, features derived from the coefficients of Fourier transform. Inasmuch
as the power spectrum is Fourier transform of autocorrelation function it can be
used to measure statistical dependence of contours. It is possible to determine
mean size, deviation of contour shape from a circle, average roughness of
contour and many others parameters [8,9].
In this work we are concentrate on the methods of shape characterization and
their classification by the Fourier descriptors and distance measures. Use of
spectral data requires the special approaches to their analysis based on the
methods of spectrum estimation. For surface damage analysis it is necessary to
have methods of classes description, their comparison and classification based
on this type of features. Particularity of given work is consideration all of
this problems from feature extraction to classifier design .
The following major steps are involved. The first is a low level image
processing for detection of the contour of the damaged area. The next is feature
extraction. We calculate Fourier descriptors for contour shape characterization.
The features are arranged as an ordered set or feature vector. The feature
vector is considered as coordinates of a point in multi-dimensional space. The
location of the point depends on the shape peculiarities of the damaged area.
For shape differentiation the distance and similarity measures were used. The
last was classification of damaged particles. Here a location of test point or
cluster of points in the feature space examined and compared with the classes’
representatives. The hierarchical and k-means clustering algorithms and
multi-dimensional scaling were used for classifier design. The performance of
classifier was evaluated by using real set of damaged area.
Phase I technical objectives
This project is dedicated to development of diagnosis system of the metal
surface damage or deformation and compatibility solutions to enable large step
improvements of productivity due to protect surface quality claim in advance and
save production loss by automatic surface quality control system. The basic
technical objective of this project phase I is identification and explanation of
major compatibility issues between surface condition of any kind of metal and
it’s digital image analysis by morphological interpretation of color, shape, and
texture of surface damage in real time. Specifically, Phase I is to demonstrate,
by way of simulation tests of image capture for surface defects, the tradeoffs
associated with conflicting requirements of surface condition, for surface
deformation from high stress or temperature of production process, surface
defect from corrosion inhibition and materials fatigue life, wear resistance and
corrosion resistance with lubricant for process of milling, forging, pressing,
Phase I work plan
In order to achieve the above-mentioned Phase I objectives, the following tasks
will be performed, each of which will be described in detail below:
Task 1. Development of image captures system and database establishment
procedure for surface damage.
Task 2. Development of data expert system with data query management.
Task 3. Development of interpretation way of morphological defects for metal
Task 4. Selection of classification method of defined images.
Task 5. Establishment of automatic on-line prognosis system for surface damage.
Task 6. Preliminary testing of automatic on-line system in connection with main
Task 7. Preparation of final report and phase II investigation plan.