Ðóññêèé   Óêðà¿íñüêà
DonNTU   Master's portal

Abstact of master's work

Contents

Ââåäåíèå

There are a lot of systems in the modern information system’s world, the main functionalities of which are identification process of materials on the image or form’s recognition on the image. The information system that will be developed during master’s work is such type of system. This one is for facilitate the analytical processing of images for scientific laboratory workers. These laboratory workers are specialized on ion-plasma technology, for example, they create coatings by vacuum arc deposition in Science Education Center “Ion-plasma technology” named by N. E. Bauman.

1. The general information about master’s work theme

1.1 The relevance of the theme

Cathodic arc deposition or Arc-PVD is a physical vapor deposition technique in which an electric arc is used to vaporize material from a cathode target. The vaporized material then condenses on a substrate, forming a thin film. The technique can be used to deposit metallic, ceramic, and composite films.

Cathodic arc deposition is actively used to synthesize extremely hard film to protect the surface of cutting tools and extend their life significantly. This is also used quite extensively particularly for carbon ion deposition to create diamond-like carbon films. Because the ions are blasted from the surface ballistically, it is common for not only single atoms, but larger clusters of atoms to be ejected. Thus, this kind of system requires a filter to remove atom clusters from the beam before deposition. Filtered Cathodic arc can be used as metal ion/plasma source for Ion implantation and Plasma Immersion Ion Implantation and Deposition [1].

Laboratory workers observe coatings using different kinds of microscopes and determinate the microdrops’ square manually using the photo editors for estimating of spraying quality. These operations are time-consuming and the process of determining the coating quality lasts for a long time. Therefore, employees need IS to facilitate such process.

The relevance of developing this type of system is that there are a few ones nowadays. Those systems can identify microdrops in the microscope’s image of coating with maximum precision. But it’s quite necessary to notice a fact that there is some program software for powder products’ recognition, such as “Cinderella” [2], and for metallic materials – VT-1600 [3]. Unfortunately, those systems do not consider the features of substances, projection of which will be explored in this work.

Paying attention to what have been written above, the developing program application is going to be pretty effective and efficient solution for the future users of the system..

1.2 Goals and tasks of research

The main purpose of the master's work is creation of the information system for analytical processing of images to automate the process of determining the coating quality. That process includes identification of droplets in the coating and determination their characteristics to facilitate the scientific of image processing analysis for laboratory workers specializing in ion-plasma technology.

The information system for image processing is a complex solution to determine the microdroplets in the coating and calculate their physical characteristics. Images are taken by using confocal and languidly force microscope.

As a result, it is going to build very convenient user’s interface and to reach the goal. The goal is that the count of a right recognized microdroplets in the coating will be more than 50 percent. In addition, there is a need to solve following tasks:

  1. Reduce the image processing time at the expense of using modern algorithms of form’s recognition;
  2. Increase a count of the right recognized microdroplets in the coating.

Object of research: process of identification of microdroplets in the coating, which was deposited by vacuum arc deposition.

Subject of research: microdroplets identification’s methods on the images..

1.3 Scientific novelty of the results

The information system, which is using by laboratory workers nowadays, cannot identify microdroplets in the coating, determine the center of the circles of substances and cut the other material’s microdroplets off. The other droplets might be in the image. There can be dust.

If look closely to the picture person can find out, that figures with form of the circle have the highlight in the inner part, this is the hallmark whether there is a microdroplet or a speck of dust.

Because there is no such functionality in current system, the decision was made to design and develop software, which allow finding microdroplets in the coating. This will greatly simplify the identification drops process. Drops presently are determining manually via external attachments and conversions from pixels to nanometers and back.

Consequently, the novelty of IS development is in the next things: this is first time when decision support system will be developed for the determination of the quality of coverage, which deposited by vacuum arc deposition, and when algorithm of identifying process will be established.

1.4 Application development

This information system can be used by for scientific laboratory workers. These laboratory workers are specialized on ion-plasma technology.

IS developing guarantees the microdroplets recognition quality, convenience of using, remote access to data. The perspective of evolution and system modernization are in that fact it’s planning to create coating 3D model with marked and recognized microdroplets of materials.

2 Overview of Research and Development

2.1 Review of international sources

ÎDescription and comparison of pattern recognition methods were discussed in the T.T.CH. Buoy’s work [4], various ways and some advice how to process image effective were described in the book of R. Gonzalez [5]. The idea about data’s or repeating operation’s parallelization can appear during image processing. The group of scientists (the head was N. E. A. Halid) wrote about parallelization in the article [6]. Integrated edge detection method and using of mathematical morphology were highlighted in the work of K. P. Huang [7]. We can learn about image binarization from the book, which was written by B. C. Mors [8].

2.2 Review of national sources

Description of Powder Coating by vacuum-arc was presented in a scientific paper by A. Pirogov [9], îit was told about object recognition using Otsu method in the article of J. Koslov [10]. It is quite good paper about an action sequences in pattern recognition by A. Maltsev [12].This article has had an impact on choice of the algorithm. That algorithm will be used in the master’s work.

3 Procedure of identifying the object

It proposed to use method of the selection borders, which based on Sobel’s operator, and image binarization for solving tasks. Chosen action’s sequence is presented at the fig 3.1

Chosen action’s sequence

Figure 3.1 – Chosen action’s sequence

It planned to use method Otsu for binarization of image. In computer vision and image processing, Otsu's method, named after Nobuyuki Otsu, is used to automatically perform clustering-based image thresholding, or, the reduction of a graylevel image to a binary image.

The algorithm assumes that the image contains two classes of pixels following bi-modal histogram (foreground pixels and background pixels), it then calculates the optimum threshold separating the two classes so that their combined spread (intra-class variance) is minimal, or equivalently (because the sum of pairwise squared distances is constant), so that their inter-class variance is maximal. Consequently, Otsu's method is roughly a one-dimensional, discrete analog of Fisher's Discriminant Analysis.

The extension of the original method to multi-level thresholding is referred to as the Multi Otsu method.

In Otsu's method we exhaustively search for the threshold that minimizes the intra-class variance (the variance within the class), defined as a weighted sum of variances of the two classes:

Similarly, you can compute and on the right-hand side of the histogram for bins greater than t. The class probabilities and class means can be computed iteratively. This idea yields an effective algorithm:

  1. Compute histogram and probabilities of each intensity level
  2. Set up initial
  3. Step through all possible thresholds t=1 maximum intensity
    a) Update
    b) Compute
  4. Desired threshold corresponds to the maximum image height="25px" src="images/ref_formula7.jpg">
  5. You can compute two maxima (and two corresponding thresholds).
  6. Desired threshold = [12].

For edge detection – Sobel method. The idea of this method is based on imposition two rotation masks on each point of image . These masks are two orthogonal dimensions 3x3 matrix, which are presented in following table 3.1.

Table 3.1 “Rotation masks of Sobel method”

1st matrix 2nd matrix
-1 0 +1 +1 +2 +1
-2 0 +2 0 0 0
-1 0 +1 -1 -2 -1

These masks detect the edges. The edged locate vertically and horizontally on the image. There is a possibility to get the estimate of the gradient per every dimensions Gx, Gy, when it is using separation imposition of these two masks. The final value of the gradient is determined by the formula: G=v(G_x^2+G_y^2) [4].

The criteria of that recognized object is microdroplet is the highlight in the center of the circle. Highlight is a group of horizontally and vertically located pixels. Values of RGB of those pixels differ sharply from pixel-neighbors. It is a group of contiguous pixels with values of R=0, G=0, B=0 on binary pictures of coating with microdroplets.

During first launching of the system there will be the initial configuration of system parameters. It includes values variation of threshold in Otsu method, correction of coefficients of "sliding" window and dimensions in a case if the object recognition quality is far away from the desired one.

The percentage of correctly classified objects (in this work such objects are microdroplets) will be calculated based on the image, on which expert allocated manually microdroplets in the coating.

After the next run of the program if the quality of recognition satisfies users, the final stage will be to identify physical parameters of the droplet.

4. Model of developed system.

The process of user system interaction begins with his/her registering therein. User fills necessary fields. Then system analyses: whether this user’s data are in DB system or not. In a case, if there is no such user system suggest to register in it or enter another data in necessary fields (if person want to continue use the application).

If registration succeed, then system define the user’s rights and roles.

If system finds that data in DB and user. User has possibility to load pictures for analysis. File will be filtered after loading coating projection image. Filtering helps to detect interesting areas on image. At the filtering level image analysis is not performed. But objects, which stay after filtering, can be reviewed as areas with special characteristics. These areas may be contenders for the "title" microdroplets. The filter at the initial phase will be the binarization. Filtering provides a set of useful data for processing. But, as general, there is no opportunity take and use the data without processing.

Phase of logic processing of filtering results comes after last step. The edge analysis is doing. Selected areas, which have size less than 10 nm, will be cut off. If there no highlight inside of circle, this area will be also cut off.

Then information system calculates the center of the circle, radius, square of this circle.

After determination of the physical parameters of the microdroplets in the coating there is a possibility to make a conclusion about coating quality. Conclusion is based on percentage. System calculate square of hole pictured (1 cm is equal to 37.76 pixels). Then all squares of all identified objects sum. Percentage is calculation as the ratio of the summed square of microdroplets in an entire area of image.

Modification of the incoming coating picture as a result of the information system is shown in Fig. 4.1.

Modification of the incoming coating picture as a result of the information system

Figure 4.1 – Modification of the incoming coating picture as a result of the information system (animation — 4, delay — 1000 ms, size — 240 Kb)

Conclusion

There was hold analysis of the edge detection methods and of the input image binarization in this master’s work for creation information system of identification microdroplets in the coating. As result, the sequence of methods and actions was pick up. It provides the highest performance of information system. The algorithm of identification and methodic of recognition were also developed.

The main prerequisites are that a lot of already existing software do not recognize microdroplets in the right way, because those one were developed for another type of substances.

At the present time the master’s work is still writing. It is not over. There will be opportunity to familiarize with master’s work at December of 2015, when it will be at the final phase.

List of sources

  1. L. P. Sablyev, Yu. I. Dolotov and others. «Apparatus for vacuum-evaporation of metals under the action of an electric arc» [Electronic resource] // Access mode: http://patft.uspto.gov/netacgi//...
  2. Cinderella - system analysis of the size and shape of solids [Electronic resource] // Access mode: http://www.mallenom.ru/cinderella.php
  3. The optical particle size analyzer by digital microscopy VT-1600 [Electronic resource] // Access mode: http://www.analizator.su/BT-1600.php
  4. Bui T.T.T., Spitsyn V.G “Analysis of methods of digital images edge detection ” [Electronic resource] // Access mode : http://www.tusur.ru/filearchive/reports-magazine/2010-2-2/221.pdf
  5. Gonzalez R. Digital Image Processing / R. Gonzalez, R. Woods . – M.: Technosphera, 2005. – 1070 p. [Electronic resource] // Access mode : http://www.technosphera.ru/files/book_pdf/0/book_311_455.pdf
  6. N.E.A.Khalid, S.A.Ahmad, N.M.Noor, A.F.A.Fadzil and M.N.Taib «Parallel approach of Sobel Edge Detector on Multicore Platform» – INTERNATIONAL JOURNAL OF COMPUTERS AND COMMUNICATIONS - p. 236-244 – [Ýëåêòðîííûé ðåñóðñ] // Ðåæèì äîñòóïà: http://www.naun.org/main/UPress/cc/17-303.pdf"
  7. . Huang C.P. An Integrated Edge Detection Method Using Mathematical Morphology / C.P. Huang, R.Z. Wang // Pattern Recognition and Image Analysis. – 2006. – Vol. 16, ¹ 3. – P. 406–412. – [Ýëåêòðîííûé ðåñóðñ] // Ðåæèì äîñòóïà: http://link.springer.com/article/10.1134%2FS1054661806030102?LI=true
  8. Bryan S. Morse «Thresholding» [Ýëåêòðîííûé ðåñóðñ] // Ðåæèì äîñòóïà: http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/MORSE/threshold.pdf
  9. Ïèðîãîâ À. À. «Ïîäñèñòåìà ÀÑÒÏÏ íàíåñåíèÿ âàêóóìíî-äóãîâûõ ïîêðûòèé ñ çàäàííûìè ñâîéñòâàìè íà ïîâåðõíîñòü ñòåêëà» [Ýëåêòðîííûé ðåñóðñ] // Ðåæèì äîñòóïà: http://www.dissercat.com/content/podsistema-astpp-naneseniya-vakuumno-dugovykh-pokrytii-s-zadannymi-svoistvami-na-poverkhnost
  10. Êîçëîâ Äæ. «Îáíàðóæåíèå îáúåêòîâ ìåòîäîì Îöó» [Ýëåêòðîííûé ðåñóðñ] // Ðåæèì äîñòóïà: http://habrahabr.ru/post/112079/
  11. Ìàëüöåâ À. «Ïàðó ñëîâ î ðàñïîçíàâàíèè îáðàçîâ» [Ýëåêòðîííûé ðåñóðñ] // Ðåæèì äîñòóïà: http://habrahabr.ru/post/208090/
  12. «Ìåòîä Îöó» [Ýëåêòðîííûé ðåñóðñ] // Ðåæèì äîñòóïà: https://ru.wikipedia.org/wiki/%D0%9C%D0%B5%D1%82%D0%BE%D0%B4_%D0%9E%D1%86%D1%83