WWW.DISSERS.RU


...
    !

Pages:     | 1 |   ...   | 38 | 39 || 41 | 42 |   ...   | 54 |

the element of more detailed partition will be the result of the image retrieval. In simplest situations it is necessary to fulfill a single-stage clustering, i.e. to optimize retrieval under worst-case conditions we have to ensure (2) [ ]}+ [ ]) N2 = card{ X max(card X min.

At the multilevel clustering the repeated clusters search inside of already retrieved clusters is fulfilled and only on the last step required image is searched by complete enumeration. Let us assume that the cluster [X(i-1)]p is selected on (i -1) level of hierarchy from a condition (y, [X(i - 1)] ) min, q = 1, card{[X(i - 1)]}, i.e.

q [X(i-1)]p = = [X(i)]1 [X(i)]2 [X(i)]p where for any k and l the equality [X(i)]k [X(i)]l = holds. Then the minimization of matches amount is reduced to the clustering with the goal function m [ ]p, (3) N3 = {card [X(i)]p, i x [X(i-1)]p, (i-1)} + max (card X(m-1) (m-1)) min, i=-1 ( ) where m is a number of hierarchy levels, [X(0)]1, 0 = X. The method of search (1) was proposed in [4], the ( ) solution of problem (2) was offered in [5], searching of (3) one can see in [6].

Content of an image may be often summarized by a set of homogeneous regions in appropriate feature space.

Therefore, there is a great need for automatic tools to classify and retrieve image content on the base of segmentation.

Segmented images are formed from an input image by gathering its elements into sets likely to be associated with meaningful objects in the scene. That is, the main segmentation goal is to partition the entire image into disjoint connected or disconnected regions. Unfortunately, the effectiveness of their direct interpretation depends heavily on the application area and characteristics of an acquisition system. Possible high-level region-based interpretations are associated with a priori information, measurable region properties, heuristics, plausibility of computational inference. Whatever the case, often it is necessary to have dealings with a whole family of partitions and we must be able to compare these partitions which are produced by a variety of segmentation algorithms. At least splitting and merging techniques make us to match segmentation results which ultimately may be corresponded to indirectly images comparisons.

For region-based similarity analysis novel approaches are required since usually early processing scheme consists of following steps: images are segmented into disjoint (or weakly intersecting) regions, features are extracted from each region, and the set of all features is used for high-level processing. It should be emphasized that quite often simultaneous processing of partitions or coverings is wanted to produce reliable true conclusion.

In this connection we propose and vindicate a new metric providing all partitions (and consequently images) matching n (4) (P,Q) =| XkYl || Xk Yl | m l=k =where denotes a symmetric difference, P = {X1,X2,,Xm}, Q = {Y1,Y2,,Yn} Xk,Yl D, D is a field of view (generally, arbitrary finite signal or feature space with found partitions ). Note, these partitions are segmentation results, representing pairwise disjoint family of nonempty subsets whose union is the image and each subset may contain required target, may belong to a carrier of object image or may be a part of that. From this follows to provide possibilities of reliable low-level feature selection and reasonable semantic concepts accommodation often it is necessary to analyze partition collections.

Fourth International Conference I.TECH 2006 Fig. 1. Examples of input images Fig. 2. Multithresholding segmentation of images shown in fig. Fig. 3. Examples of image and partitions matches (query a) and b) correspondingly) The analysis of experimental results has shown that the application of partitions as features provides a sufficient relevance at access to an image in database with queries ad exemplum. Figures 1 and 2 illustrate images and partitions which were compared by means of traditional and proposed metrics. Examples of dependences, query image and its partition are shown in fig. 3. We see comparability of obtained results for Euclidean metric and distance function (4). The reliability of matching can be increased by an intellectual processing (via relations analysis of region-based models) which provides conditions for entirely correct and complete segmentation.

Conclusion An intensive experimental exploration with the collection of histologic specimens images with final goal classification as an aid in cancer detection vindicates the efficiency of proposed metric.

Bibliography 1. Mller H., Mller W., Squire, D.McG., Marchand-Maillet S., Pun T. Performance Evaluation in Content-Based Image Retrieval: Overview and Proposals. Pattern Recognition Letters, Vol. 22, 2001, pp. 593-601.

2. Li J., Wang J.Z. Automatic linguistic indexing of pictures by a statistical modeling approach. IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 25, 2003, pp. 1075-1088.

3. Kster T., Wendt V., Sagerer G. Comparing Clustering Methods for Database Categorization in Image Retrieval.

Lecture Notes in Computer Science, Springer-Verlag, Berlin Heidelberg, Vol. 2781, 2003, pp. 228-235.

Software Engineering 4. Mashtalir V.P., Yakovlev S.V. Point-Set Methods of Clusterization of Standard Information. Cybernetics and Systems Analysis, Kluwer Academic Publishers, Vol. 37, 2001, pp. 295-307.

5. Kinoshenko D., Mashtalir V., Yegorova E. Clustering Method for Fast Content-Based Image Retrieval. Computer Vision and Graphics, K. Wojciechowski et al., eds., Computational Imaging and Vision, Springer, Vol. 32, 2006, pp. 946952.

6. Kinoshenko D., Mashtalir V., Yegorova E., Vinarsky V. Hierarchical Partitions for Content Image Retrieval from LargeScale Database. Machine Learning and Data Mining in Pattern Recognition, P. Perner., A. Imlya, eds., Lecture Notes in Artificial Intelligence, Springer-Verlag, Vol. 3587, 2005, pp. 445-455.

Authors' Information Mashtalir Vladimir Doctor of Technical Sciences, Professor of Computer Science Department and Dean of Computer Science Faculty, Kharkiv National University of Radio Electronics, Lenina ave, 14, Kharkiv, Ukraine, 61166, e-mail: mashtalir@kture.kharkov.ua Shlyakhov Vladislav Candidate of Technical Sciences (equivalent Ph.D.), Senior Researcher of Computer Science Department, Kharkiv National University of Radio Electronics, Lenina ave, 14, Kharkiv, Ukraine 61166.

DEVICE FOR COUNTING OF THE GLASS BOTTLES ON THE CONVEYOR BELT Ventseslav Draganov, Georgi Toshkov, Dimcho Draganov, Daniela Toshkova Abstract: In the present paper the results from designing of device, which is a part of the automated information system for counting, reporting and documenting the quantity of produced bottles in a factory for glass processing are presented. The block diagram of the device is given. The introduced system can be applied in other discrete productions for counting of the quantity of bottled production.

Keywords: device for counting, automated information system Introduction In all discrete productions it is needed the ready production to be counted as well as reporting and documenting of the received data. In the present paper a device for counting the quantity of the produced glass bottles, moving on conveyor belt and which is designed by the authors is presented. It is a part of the automated information system for reporting and documenting of the ready production in a factory for glass processing [Draganov, 2006].

The information system has to meet following requirements: collecting data for the ready production, moving in one direction on the conveyor belts; archiving the data for each shift; reporting the quantity of the production for a shift (eight hours).

Different company developments of production counting systems are known [Solid Count, 2006; Fast Counts, 2006; Patent 0050111724, 2005]. One of them is the system SolidCountTM [Solid Count, 2006], which is designed for an automatic collecting of data for the ready mixed (of different kinds) production from a single production line, reporting the quantity of the production and receiving statistical data for the production in real time. The system Fast CountTM [Fast Counts, 2006] serves for: collecting data from several lines; reporting of the quantity of the production in different formats; monitoring of the productivity; archiving of the data; statistics and diagnostics in real time. For counting of the ready production a method and apparatus for counting is suggested in [Patent 0050111724, 2005]. The data for the ready production are received by comparison between the image of the product on programmable zoned arrays of light sources and photo detectors and known images.

Fourth International Conference I.TECH 2006 The software and hardware products, which are considered, are of general use. They are expensive, very complicated and less reliable. These disadvantages are avoided in the system for counting, reporting and documenting of ready production, moving in unidirectional way on four conveyor belts as well as the entire production of the factory for glass from the four conveyor belts. The system is developed by the authors and it is introduced in a factory for glass.

Structural Diagram of the Automated Information Systems The structural diagram of the automated information system in the factory for glass is depicted in Fig.1.

Fig.1 Structural diagram of automated system for counting of bottles on conveyor belt Each of the four input conversion devices (ICD) feeds an electric impulse to the device for counting control and indication (DCCI) when a ready production unit passes the conveyer belt in front of the input conversion device.

In DCCI information about the quantity of the impulses, which have come from the four ICD, is gathered. On the basis of this information the necessary signals for control of the indications I1I5 are depicted. In the presence of danger of overflow of any of the counters, registering the input impulses, DCCI sends a signal for overflow ( O) to the device for printing control (DPC). The last also receives information for the state of all counters in DCCI (Q).

DPC gives a command to the printing device (PD) for printing the results and after that to DCCI a command to clear the counters (R). The printing with consequent clearing is also accomplished by external signal from an operator through clearing button, lying on the command panel, which is a part from the DCCI, at the end of the shift. In case of power failure DPC saves the current information and after restoring the electricity supply the necessary commands for printing and clearing are passed to PD.

Scheme Solution of the Device for Counting of Bottles To receive reliable information for the quantity of the produced glass bottles it is necessary each input converting device from the automated information system for reporting and documenting of the quantity of ready production to be designed. The device has to meet the following requirements: to convert the information for the number of the glass bottles, which move on the conveyor belt separately or in groups in electrical impulses with TTL level in contactless way; the number of the electrical impulses to correspond strictly to the number of the passing glass bottles and errors, caused by bottles, which are contiguous one to another or by the uneven optical density of the glass from which the bottles are made or by vibrations of the conveyor belt have to be expelled; the device to be Software Engineering simple and cheap at most and with high reliability of the scheme solution; the construction to be with high mechanical stability and manifacturability.

The main goal of the work is to design a device, which meets the attached requirements and free of the indicated disadvantages.

Devices for counting of objects, based on electro-contactable, capacitive, inductive and other principles are known. One of the most perspective one is the photo-converting principle, which has following advantages: broad field of application; contactless way of operation; high reliability and long exploitation time; high promptitude; low feeding voltages and small consumption of electrical power; broad temperature range of operation; possibility for miniaturization and integration and etc.

The photo-converting devices frequently operate in a mode of transmission [Bergmann, 1980], in which the counted objects cross and modulate a ray, emitted from light source to a light receiver, situated on the other side of the object. There is a possibility for operating in another mode mode of reflection [Bergmann, 1980], in which the light source and the light receiver are situated on one and the same side of the moving object, reflecting directly or diffusely part of the light, emitted by the light source to the light receiver. An operation in a mode of autonomous emitting [Bergmann, 1980] at which the object itself is a light source is possible.

The photo-converters may operate with unmodulated and modulated light [Bergmann, 1980]. The schemes of the photo-converters with unmodulated light are simplified but they are adversely influenced by the disturbing light daylight or artificial, emitted by other sources of light. The photo converters with modulated light are protected from the influence of the disturbing light in a high degree, but their scheme solution is complicated and expensive.

In the designed device the photo converting principle of operation, based on mode of transmission of the unmodulated light is used. Thus a simplified scheme solution is obtained.

The disadvantages of principle of the devices operating with unmodulated light are not substantial in the concrete case as the application of the device to be designed is characterized by a small distance between the light source and the light receiver and lack of parasitic lighting. For the purpose an appropriate construction is developed.

The possible errors, caused by vibrations of the conveyor belt and by the uneven density of the bottles may be avoided by transmission of light ray at the height of the mouth of the bottles. But even in this case the light ray is discontinued repeatedly when a single bottle is passing and the number of the obtained output impulses is arbitrary.

Pages:     | 1 |   ...   | 38 | 39 || 41 | 42 |   ...   | 54 |



2011 www.dissers.ru -

, .
, , , , 1-2 .