ISA District 12 students sections

St.Petersburg State University of Aerospace Instrumentation students section (Russia)

The system of the warehouse registration for book trade

Vitaly Sedunov

This program is intended for the registration of document circulation in a book warehouse. The application has the well organized and easily modified data structure, at the expense of that it can be easily adapted under trade in other goods, for example, chancery. A lot of Russian users know lacks of existing systems of the warehouse registration computer systems: slow speed and unreliability of operation at the large assortment, impossibility of moving of the goods on a trade point or shop, passing a central warehouse. The given program liquidates these lacks and has a lot of other advantages: Besides processing with documents the given program allows to make the cash orders, to keep account of the accepted bills - invoices, to keep in database the various prices for one name for different shops, but to view total statistics as under one name, possibility of direct operation with e-mail, possibility of export of the data in other accounting systems, distributed in Russia, has various opportunities of statistical data processing, possibility of export of the data in Microsoft Excel.

Interface of program for tests and debugging of plane brake systems

Sergey Zhukov, Natalia Ovchinnikova

This applied program is intended for work in structure of the stand Test Bench of tests and debugging of plane brake systems with the block of management of process of braking in a contour of management.

The stand provides an opportunity of tests plane's antiblocking systems (ABS) in all modes of braking in various conditions of operation, simulates actions of bodies of management ABS and other onboard systems of the plane. It provides also opportunity of registration and processing of parameters during tests of system in various modes.

When starting the program a panel with three inserts is displayed:

Initial conditions;

Simulation;

Simulation Results.

Clicking the left button of the mouse makes selection of an insert by corresponding insert.

initial conditions insert provides input of main initial data of simulation mode. It contains six panels with initial conditions of simulation and three keys:

<OK> - confirmation of input of initial conditions;

<Cancel> - cancellation of input of initial conditions;

<Help> - call for reference.

For data input one-line input editors are used to which input focus should be delivered. To do that click left button of the mouse on corresponding window or <Shift> key to switch to input window (combination of <Ctrl+Shift> keys to move in opposite direction). Data are entered by keyboard.

Data on runway condition on "External conditions" panel and type of the wheel on "....(. Wheel" panel are entered through editor with the list. They can be entered by keyboard, e.g. - "?.5.A.B.@.>." for "Runway condition" field. Key (t) on the right gives access to the list where one can select one of possible types of data. If data entered is not available the user is notified about this and default data are set up.

Clicking left button of the mouse by the radio knob you need makes selection of one of the data "Height of an Airfield "or" Air Pressure on "External Conditions" panel. In this case the second datum becomes inaccessible for input and cannot be used in the program.

Digital adjuster enters the quantity of engines on "Aircraft engine" panel.

Angles of release of flaps and interceptors are entered on "Aerodynamics" panel with the help of respective radio knobs.

"Simulation" insert provides execution of simulation. It displays time diagram of horizontal component of airframe's speed thus visualizing simulation process and three buttons:

<Run> - execution of simulation;

<Stop> - stop of simulation;

<Help> - call for reference.

"Simulation Results" insert provides view of simulation results represented by diagrams and charts. It displays time diagrams of processes in the systems and seven buttons:

<Table > - review of simulation results represented by a table;

<Save> - saving of simulation results using standard Save-dialogue in Windows'95 (default - in file with preset user's name and 'nba' extension);

<Select> - selection of a set of processes subject for review;

<View> - review of simulation results represented by diagrams;

<Open> - review of previously saved simulation results with the use of standard Open-dialogue in Windows'95 (default - in file with preset user's name and 'nba' extension);

<Exit> - exit from the program;

<Help> - call for reference.

When viewing simulation results use switches to select data of interest to you (<Select> key), which are placed in respective inserts. Data presented in

"Default" insert are given as default.

View of diagrams (<View> key) can be adjusted to the value of visible time interval with the help of digital adjuster. In case part of the diagram is not visible moving the mouse horizontally with pressed right button you can drag it without changing time scroll. To increase the scale of a diagram you should mark the fragment you need by pressing left button of the mouse in its left upper corner and releasing it in the right bottom corner. To return to the initial scale press the left button of the mouse and move the mouse to the left upward.

To view numerical values of active diagrams "Values" switch must be activated. Cursor will appear in diagram field and next to the "legend" - numerical values for respective time moments. Moving diagrams in the way described above you can get values of coordinates for time moment of interest to you.

The Texture Classification System Model

Sergey Kot

Texture analysis is an important step in many visual tasks, such as object recognition, scene segmentation and shape perception. The model described in this work is based on representing textures in frequency and orientation space with using the non-parametric learning scheme for classification [1] and the shift distortion model [3]. Classification procedure is maid according to selected decision making rule and based on a set of features parameters obtained after learning stage.

Textures images are usually easy identifiable by a human vision, but no simple mathematical definition captures all aspects of the very diverse texture family. All texture family is generally divided into two main types: micro (structured or oriented) and macro (unstructured or non-oriented) textures [2] (see fig.1). Many existing texture descriptors are based on the point of view that texture is a regular pattern composed of repeated primitive elements. Most natural images do not generally show such regularity. Some prototypical examples of unstructured textures, which are known as random macro textures are ceramic, marble and granite images. The placement of the primitives within these is purely random and highly irregular.

Figure 1. Oriented and non-oriented textures (left and right respectively).

For the more structured textures, geometrical models are frequently used to define primitives and the spatial arrangement between them. Learning stage of classification procedure can be used for adaptation the geometrical templates to the texture data, as well as to extract an informative set of rules for the discrimination task. Automatic inspection needs to be able to define a priori a good set of primitives and placement rules in order to characterize the textured input. Thus, deviations from the texture primitives should be easily identifiable and, more importantly, measurable. One of the widely used methods in texture analysis is the computation of gray-level co-occurrence matrices. These matrices provide a full representation of the second-order gray-level statistics of textured images. Some problems associated with the co-occurrence matrices are that they require much computation, as many co-occurrence matrices need to be computed, for varying neighborhood sizes.

Stochastic models, for instance the Markov Random Field models, can be used to represent unstructured or stochastic textures. These models are defined a priori and capture the spatial contextual information in an image with the assumption that the brightness at each pixel is random value and depends on the brightness of only its neighboring pixels. Learning is used to estimate the model parameters, via an optimization process. Stochastic models can describe many types of textures very well, but textures can be found which fall outside the model definition.

Other methods compute texture features by filtering the image and using these filtered characteristics in the classification task. Spatial domain filters which include simple edge masks and more complicated masks which are based on spatial moments, can be used to filter images. Non-parametric classification scheme [1] is used for the texture classification task. Learning procedure is used to identify the important characteristics of the input image and to extract the appropriate set of features for the task after filtering the texture image.

As we know, each learning algorithm of a classification procedure has advantages and disadvantages, but on most problems of these methods will prove successful in predicting the class output at a performance approaching the Bayes rate. Classifier utilizes an information theoretic rule-based learning scheme, which provides for probabilistic classification, according to the standard non-parametric nearest neighbor rule [1].

The system includes a feature extraction stage followed by a learning stage. The feature extraction stage of the system consists of a Gabor decomposition scheme. Orientation and spatial frequency responses are extracted from local areas of the input image and the statistics of the coefficients characterizing the local areas (or `windows') form representative feature vectors. The eight filters are adequate cover the 360 degrees of orientation space with more than 99% accuracy [1]. Feature-vectors are formed from the extracted power of filters set associated with each image local-area. It is needed to discriminate texture pairs with identical mean brightness and identical second-order statistics. The estimation of the classification system model showed that the model has high-percentage classification rates (96 % and more at 32x32 pixels of image windows) for wide variety of textures. This estimation is performed with using the mathematical model of texture images shift distortion [3].

References

1. H. Greenspan, R. Goodman, R. Chellappa, and C. Anderson. Learning texture discrimination rules in a multiresolution system. IEEE Transactions on Pattern Analysis and Machine Intelligence, 16(9):894-901, September 1994.

2. R.M. Haralick. "Statistical and structural approaches to texture." IEEE Proceeding, 67(5):768-804, 1979.

3. Mathematical Models and Criteria of Decision Making in Images Identification, Kot S., // EUROXCHANGE (ISA European Region paper competition - ERSPC), ¹ 2 (Spring) 1999, Netherlands, pp. 2 - 21

 

THE MEASURING SOFTWARE FOR ACCEPTANCE TEST SYSTEM FOR IR-CCD BASED ON SCHOTTKY-BARRIER DIODES

Alex Sergeev

The measuring software for acceptance test system for IR-CCD based on Schotky-barrier diodes in manufacturing conditions is considered.

The software is developed for an operating system Windows 9x, with usage BCB 3.0 CSS. The basic performances and flowchart of a stand, with the description of correlation between its constituents are adduced.

The software includes following modules: the control module by operational modes a IR CCD; a module of image entry created IR - CCD in real time, module of image output on the display screen of the computer; a module of statistical processing both maps as a whole, and its fragments; a module of documenting of the data and images.

The methods of implementation of statistical image processing are described, the examples of static and dynamic versions of performance of the obtained statistical data are adduced.

The ways of image processing facilitating realization of measurements (subtraction of dark frame) are considered.

The examples of the real images illustrating problem solving put at software engineering and considered in the report are adduced.