Wyniki wyszukiwania - BazTech

1

Application of genetic algorithm for image searching on example of STA

Polaczek J.

Journal of Theoretical and Applied Computer Science

|

2015

|

Vol. 9, nr 1

9--16

EN

Image tracking and recognition is one of the key data processing mechanism for humans. Because of that the implementation of image recognition mechanisms is visible in many branches of science. Mathematical approach to the problem allows to propose many different solutions, which effectiveness depends on the way of defining the problem. The following article shows one of the possibilities of applying genetic algorithm to solve the problem of image tracking from defined set of digital images. The aim of the article was to analyse the effectiveness of used genetic algorithm for image tracking. The article states a thesis, that it is possible to use a genetic algorithm utilizing Query By Image Content mechanism with an acceptable (by W3C standards) return time. Designed by author, variation of genetic algorithm working on defined set of digital images, was put under evaluation. Based on performed research, the effectiveness and speed of image tracking was evaluated with use of correlation coefficient and measures of algorithms time of work. Performed research showed the possibility of implementing the genetic algorithm in similar image tracking, with results exceeding the expectations of computer users.

2

Zastosowanie zmodyfikowanego przekształcenia Wignera-Ville’a do analizy sygnałów zmodulowanych

Kraszewski T.

Elektronika : konstrukcje, technologie, zastosowania

|

2014

|

Vol. 55, nr 6

41-44

PL

Współczesna analiza sygnałów oparta na standardowym podejściu do ich przetwarzania, poprzez ich czasową lub częstotliwościową analizę, nie zawsze prowadzi do satysfakcjonujących rezultatów. W wielu dziedzinach nauki i techniki coraz bardziej popularne stają się algorytmy wykorzystujące przekształcenia czasowo–częstotliwościowe dające pełniejszą wiedzę o analizowanym sygnale. W artykule zaprezentowany został algorytm wykorzystywany do analizy sygnałów zmodulowanych częstotliwościowo (ze szczególnym uwzględnieniem liniowej modulacji częstotliwości) oparty na przekształceniu Wignera – Ville’a, filtracji cyfrowej (operator Sobela) oraz detekcji linii prostych (transformata Hough’a). Rozważania teoretyczne poparte zostały wynikami badań symulacyjnych przeprowadzonych na obrazie będącym reprezentacją czasowo–częstotliwościową sygnału.

EN

A contemporary analysis of the signals based on a standard approach to their processing, through their time or frequency response analysis, does not always lead to a satisfactory results. Algorithms that use the time – frequency transformations are becoming more and more popular in many fields of science and technology because of their better description of the processed signals. The paper presents a new algorithm for the analysis of the frequency-modulated signals (linear frequency modulation in particular) using Wigner – Ville transformation, digital filtration (Sobel operator) and straight line detection (Hough transform). Theoretical considerations were confirmed by the results of the simulation experiments carried out on the time-frequency image representation of the signal.

3

Metoda rejestracji, przetwarzania i analizy obrazu do cyfrowego opisu powierzchni manekina

Lesiakowska-Jabłońska M.

Pomiary Automatyka Kontrola

|

2007

|

R. 53, nr 9

44-47

PL

Zapis powierzchni fantomów - manekinów (stosowanych powszechnie w odzieżownictwie fizycznych modeli budowy ciała ludzkiego o określo-nej uśrednionej charakterystyce wymiarowej) jest konieczny z punktu widzenia prac nad inżynierskimi metodami konstrukcji wykrojów odzieży. Potwierdzają to liczne prace z tego obszaru prowadzone przez szereg placówek naukowych i badawczych świata [1, 2, 3]. Dzięki badaniom prowadzonym w Katedrze Odzieżownictwa PŁ opracowano metodę fotogrametryczną pozwalającą na rejestrację, przetworzenie, analizę, opis cech wymiarowych i odwzorowanie 3D (three dimensional) powierzchni manekinów [4].

EN

Phantom surface representation based on computer analysis of phantom 2D images is a very challenging problem. The representation of phantom surface is necessary for development of engineering techniques for constructing shapes of clothes. The phantoms are widely used in clothing science as a model of human body. This is confirmed by large number of publications from many research institutes [1, 2, 3]. At the Clothing Department, Technical University of Łódź, the new method of digital representation of phantom surface was developed [4]. Thus optical representation method was used to obtain a digital phantom surface model. The indirect method of 3D surface measurement was proposed, where geometrical features of an analysed object were estimated based on analysis of series of its 2D images. The 2D image is a central projection of the analysed object. The most important problem is to register such a projection and then perform an appropriate analysis. To obtain a digital model of objects with size comparable to the human body, the measurement bench was assembled as shown in Fig. 1. This bench contains: analysed object (1); a projector (2) which is a source of two light half-planes (bright and dark); they are projected onto analysed object (1) and are called as "horizontal light half-plane"; a TV camera (9) as a video signal source for the computer (10); it is moved along with a projector during measurement process and it is at an angle ? in relation to light half-plane; two stepper motors (3, 4) with a driving system (5), used to move a camera (with Δζ step) and rotate a phantom (with an angle Δφ); mechanical circuit, which is used to move a camera and a projector; computer (10) used to control whole a system, equipped with two monitors (6, 7), add-on cards and specialised software; vertical reference half-plane, firmly fixed to measure a distance from the analysed object. This measurement bench was used for acquisition and analysis of images of investigated objects.

4

Application of neural networks to classification of internal damages in steels working in creep service

Dobrzański L. A., Sroka M., Dobrzański J.

Journal of Achievements in Materials and Manufacturing Engineering

|

2007

|

Vol. 20, nr 1-2

303-306

EN

Purpose: The goal of the paper is the presentation of computer assisted method for analysis of the metallographic images obtained in the scanning electron microscope (SEM) from the low alloyed steel 13CrMo4-5 elements in different states of internal damages after long time creep service. Design/methodology/approach: Investigations of the structure and morphology of internal damages resulting from creep were made by the use of light microscope and scanning electron microscope. Their topography were observed by the use of confocal laser scanning microscope. There was proposed a method based on analysis of images, shape coefficients and neural networks as a tool to evaluate the internal damage classes of materials used for the high-pressure installations elements working in creep conditions. Findings: The better efficiency of class recognition of damages developed in the material can be achieved as a combining of several methods making use of the image analysis, shape coefficients, and neural networks. Practical implications: The presented method can be use in industrial practice for evaluation and qualification of creep damage of power station boiler components operating in creep regime (e.g., steam boilers, chambers, pipelines, and others). Originality/value: Applying of the artificial intelligence method for the classification of internal damage in the steel during creep service.

5

Edge Detection and Filtering Approach Dedicated to Microstructure Images Analysis

Rauch Ł., Kusiak J.

Computer Methods in Materials Science

|

2007

|

Vol. 7, No. 2

305-310

EN

The automated analysis of images plays important role in very wide range of applications. One of such fields of interest is processing of microstructure images used to determine material grains, perform statistical calculations or prepare material model for FEM simulations. However, in all mentioned previously applications the preprocessing stage, which includes edges detection in material structure, has to be completed. The microstructure images, in most cases, consist of a set of various shapes with different sizes, what affects final efficiency and reliability of the results obtained after images preprocessing. Moreover, most of the images possess superimposed noise in form of dark spots originating from various chemical elements inside analyzed material. Unfortunately, the most of currently applied algorithms dedicated for edges detection or image segmentation e.g. Canny Detector do not cope with these problems, returning unsatisfactory results. The tests performed with application of smoothing algorithms did not success as well. In this paper authors presents the solution based on the combination of two different approaches - Particle Dynamic method dedicated for automated denoising of input images and Modified Canny Detector algorithm for edges processing. The first phase of the proposed method aims to remove additional noise from microstructure images. This functionality is performed in automated manner by minimizing set of parameters, which have to be setup before data processing. Due to this solution the time cost of data analysis is highly reduced, relieving researcher from performing many manual activities. The objective of the Modified Canny Detector algorithm is to process the image to obtain edges in the following steps: convolution filtering, edge suppression, small groups of pixels elimination. Additionally the Watershed algorithm was implemented to fulfill the edges of grains. The results obtained by application of proposed approach in comparison to other conventional methods are presented as well.

PL

Przetwarzanie obrazów mikrostruktur w celu detekcji granic między ziarnami jest wciąż trudnym zadaniem. Spowodowane jest to przede wszystkim występującym na zdjęciach szumem w postaci zarysowań lub mikro wtrąceń. Dlatego też w większości przypadków analiza zdjęć nadal wykonywana jest ręcznie, co dla dużego zbioru obrazów jest bardzo czasochłonne. Aby uniknąć tego problemu, zaproponowano podejście automatycznego przetwarzania obrazów składającego się z dwóch części tj. wykrywanie krawędzi, zaprojektowane i zaimplementowane na podstawie algorytmu Canny Detector (Ritter & Wilson, 1996) oraz filtrowania danych w oparciu o metodę cząstek dynamicznych (Rauch & Kusiak, 2005a). W wyniku zastosowania tego podejścia powstaje nowy obraz mikrostruktury z wygładzonymi obszarami ziaren oraz precyzyjnie zdefiniowanymi granicami. Osiągnięty efekt umożliwia optymalizację procesu dalszej analizy struktury materiału np.: przy użyciu algorytmu uzupełniania granic Watershed (Haris et al., 1998) czy też obliczeń statystycznych średniego rozmiaru ziaren. Artykuł przedstawia podstawowe założenia proponowanego podejścia oraz szczegóły implementacji obydwu algorytmów składowych. Wyniki przeprowadzonej analizy obrazów mikrostruktur zostały również przedstawione w niniejszym artykule.

6

Methodology of classification of internal damage the steels during creep service

Dobrzański J., Sroka M., Zieliński A.

Journal of Achievements in Materials and Manufacturing Engineering

|

2006

|

Vol. 18, nr 1-2

263--266

EN

Purpose: of this publication is to present the methodology of computer assisted method for analysis of the metallographic images obtained in the scanning electron microscope (SEM) from the elements after long time creep service. Design/methodology/approach: Stages in development of internal damage involving intercrystalline cavitation cracking were discussed and illustrated with micrographs. The method based on analysis of images, shape coefficients and neural networks was proposed as a tool to evaluate the internal damage classes of materials used for the high-pressure installations elements working in creep conditions. Findings: Combining of several methods making use of the image analysis, shape coefficients, and neural networks will make it possible to achieve the better efficiency of class recognition of damages developed in the material. Practical implications: The presented method can be use in industrial practice for evaluation and qualification of creep-damage of power station boiler components operating in creep regime (e.g., steam boilers, chambers, pipelines, and others). Originality/value: Original value of the work is applying the artificial intelligence method for the classification of internal damage in the steel during creep service.

7

Methodology of analysis of casting defects

Dobrzański L. A., Krupiński M., Sokolowski J. H., Zarychta P., Włodarczyk-Fligier A.

Journal of Achievements in Materials and Manufacturing Engineering

|

2006

|

Vol. 18, nr 1-2

267--270

EN

Purpose: The goal of this publication is to present the methodology of the automatic supervision and control of the technological process of manufacturing the elements from aluminium alloys and of the methodology of the automatic quality assessment of these elements basing on analysis of images obtained with the X-ray defect detection, employing the artificial intelligence tools. The methodologies developed will make identification and classification of defects possible and the appropriate process control will make it possible to reduce them and to eliminate them - at least in part. Design/methodology/approach: The methodology is presented in the paper, making it possible to determine the types and classes of defects developed during casting the elements from aluminium alloys, making use photos obtained with the flaw detection method with the X-ray radiation. It is very important to prepare the neural network data in the appropriate way, including their standardization, carrying out the proper image analysis and correct selection and calculation of the geometrical coefficients of flaws in the X-ray images. The computer software was developed for this task. Findings: Combining of all methods making use of image analysis, geometrical shape coefficients, and neural networks will make it possible to achieve the better efficiency of class recognition of flaws developed in the material. Practical implications: The presented issues may be essential, among others, for manufacturers of car subassemblies from light alloys, where meeting the stringent quality requirements ensures the demanded service life of the manufactured products. Originality/value: The correctly specified number of products enables such technological process control that the number of castings defects can be reduced by means of the proper correction of the process.

8

Komputerowe wspomaganie klasyfikacji wad identyfikowanych metodami radiograficznymi w odlewach ze stopów aluminium

Dobrzański L. A., Krupiński M., Maniara R., Kasprzak W., Sokołowski J.

Rudy i Metale Nieżelazne

|

2005

|

R. 50, nr 5

229-236

PL

W artykule przedstawiono komputerową metodykę klasyfikacji wad (tabl. 2), powstałych w stopach Al w trakcie wykonywania z nich elementów silników samochodowych produkowanych metodą niskociśnieniowego odlewania. Identyfikacji wad dokonano na podstawie danych uzyskanych z cyfrowych obrazów rejestrowanych metodami rentgenowskiej analizy defektoskopowej (rys. 1). Do rozwiązania tego zagadnienia wykorzystano opracowaną metodykę i związane z nią programy komputerowe do analizy obrazów rentgenowskich (rys. 2÷3), przygotowania danych wejściowych do sieci neuronowych oraz samą kontrolę jakości odlewów. Z zastosowanych w badaniach sieci neuronowych, w pracy przedstawiono wyniki klasyfikacji wad dla najlepszej sieci każdego typu. Parametry sieci o najlepszych wynikach klasyfikacji przedstawiono w tablicy 3. Zagadnienia klasyfikacyjne oceniano analizując, wyznaczoną dla danych testowych, liczbą poprawnych klasyfikacji (rys. 6÷7).

EN

In the paper a computer aided methodology of classification of defects (tabl. 2), which are formed in aluminium alloys during production of elements for car engines with the low-pressure casting method is presented. The defect identification was performed on the basis of digital recorded data registered by the use of X-ray image analysis method (fig. 1). In order to solve this problem an elaborated methodology and related computer programmes for X-ray images analysis (fig. 2÷ 3), as well as the preparation of entrance data for neuronal networks and also the quality cast control were used. The classification results of the best type of every network investigated are presented. The network parameters with the best classification results are showed in table 3. Analysing a number of correct classifications of pointed out test data (fig. 6÷7), the classifying problems are evaluated.