Wyniki wyszukiwania - BazTech

1

A homomorphic non-subsampled contourlet transform based ultrasound image despeckling by novel thresholding function and self-organizing map

Singh Prabhishek, Diwakar Manoj, Singh Shilpi, Kumar Sarvesh, Tripathi Amrendra, Shankar Achyut

Biocybernetics and Biomedical Engineering

|

2022

|

Vol. 42, no. 2

512--528

EN

The speckle noise leaves an adverse effect on ultrasound images due to which loss of information occurs. Hence this paper proposes a homomorphic Non-Subsampled Contourlet Transform (NSCT) based ultrasound image despeckling technique using a novel thresholding function, bilateral filter, and self-organizing map (SOM). The bilateral filter is utilized over the low-pass NSCT sub-band for speckle component removal and sharp features. To get better noise suppression and edge preservation, a novel thresholding function is proposed and performed over the high-pass NSCT sub-band. In the proposed method, Kohonen’s SOM is implemented as a post-processing step for deblurring purposes. The significance of the proposed scheme is also tested where it was found that using Kohonen’s SOM as post-processing works better than without post-processing. Experimental outcomes were also evaluated on real speckled ultrasound images and synthetic added speckled noisy images. The results are evaluated and compared using visual analysis and performance metrics using with and without reference images. For more critical analysis, intensity profile along a line and experts observation were also evaluated to find the performance analysis of the proposed methodology. From all experimental and comparative evaluations, it was found that the proposed approach gives better outcomes compared to similar and recent methods.

2

Evaluation of the diference in water quality between urban and suburban rivers based on self organizing map

Jiang Yaqi, Gui Herong, Li Chen, Chen Jiayu, Chen Chen, Wang Chunlei, Zhao Honghai, Guo Yan, Xu Jiying, Li Jun, Qiu Huili

Acta Geophysica

|

2021

|

Vol. 69, no. 5

1855--1864

EN

Urban rivers play an important role in maintaining the urban aquatic ecological environment, and there are bound to be differences in the water environment quality and pollution sources due to different locations of urban rivers. Therefore, this paper selects the urban river (Tuo River) and the suburban river (Bian River) in Suzhou City, Anhui, China, as the research objects. Based on the understanding of the hydrogeochemical characteristics of these two rivers, the self-organizing map is used to identify the main control factors that affect the water quality of the two rivers. The results showed that both the Bian river and Tuo river were weakly alkaline. The average content of conventional ions in Tuo river is less than that of Bian river (except HCO3 −); the water of Bian river was of Na–SO4–Cl type, and the water of Tuo river was mainly of Na–HCO3 type, with the minority was of Na–SO4–Cl type; Silicate weathering is an important source of conventional ions in the water of these two rivers; agricultural non-point source pollution is the main source of pollutants in Bian river, while Tuo river was mainly affected by natural factors, and human activities had little impact.

3

Machine learning-based analysis of English lateral allophones

Piotrowska Magdalena, Korvel Gražina, Kostek Bożena, Ciszewski Tomasz, Czyżewski Andrzej

International Journal of Applied Mathematics and Computer Science

|

2019

|

Vol. 29, no. 2

393--405

EN

Automatic classification methods, such as artificial neural networks (ANNs), the k-nearest neighbor (kNN) and self-organizing maps (SOMs), are applied to allophone analysis based on recorded speech. A list of 650 words was created for that purpose, containing positionally and/or contextually conditioned allophones. For each word, a group of 16 native and non-native speakers were audio-video recorded, from which seven native speakers’ and phonology experts’ speech was selected for analyses. For the purpose of the present study, a sub-list of 103 words containing the English alveolar lateral phoneme /l/ was compiled. The list includes ‘dark’ (velarized) allophonic realizations (which occur before a consonant or at the end of the word before silence) and 52 ‘clear’ allophonic realizations (which occur before a vowel), as well as voicing variants. The recorded signals were segmented into allophones and parametrized using a set of descriptors, originating from the MPEG 7 standard, plus dedicated time-based parameters as well as modified MFCC features proposed by the authors. Classification methods such as ANNs, the kNN and the SOM were employed to automatically detect the two types of allophones. Various sets of features were tested to achieve the best performance of the automatic methods. In the final experiment, a selected set of features was used for automatic evaluation of the pronunciation of dark /l/ by non-native speakers.

4

Pattern layer reduction for a generalized regression neural network by using a self-organizing map

Kartal S., Oral M., Ozyildirim B. M.

International Journal of Applied Mathematics and Computer Science

|

2018

|

Vol. 28, no. 2

411--424

EN

In a general regression neural network (GRNN), the number of neurons in the pattern layer is proportional to the number of training samples in the dataset. The use of a GRNN in applications that have relatively large datasets becomes troublesome due to the architecture and speed required. The great number of neurons in the pattern layer requires a substantial increase in memory usage and causes a substantial decrease in calculation speed. Therefore, there is a strong need for pattern layer size reduction. In this study, a self-organizing map (SOM) structure is introduced as a pre-processor for the GRNN. First, an SOM is generated for the training dataset. Second, each training record is labelled with the most similar map unit. Lastly, when a new test record is applied to the network, the most similar map units are detected, and the training data that have the same labels as the detected units are fed into the network instead of the entire training dataset. This scheme enables a considerable reduction in the pattern layer size. The proposed hybrid model was evaluated by using fifteen benchmark test functions and eight different UCI datasets. According to the simulation results, the proposed model significantly simplifies the GRNN’s structure without any performance loss.

5

Sztuczne sieci neuronowe ANN : sieci Kohonena

Iljaszewicz Paweł

Biuletyn Naukowy Wrocławskiej Wyższej Szkoły Informatyki Stosowanej. Informatyka

|

2018

|

vol. 8, nr 1

34--39

PL

Artykuł omawia sztuczne sieci neuronowe (ang. ANN- Artificial neural networks). Jedną z odmian są sieci Kohonena zwane Mapą Samoorganizującą (ang. SOM – Self Organizing Map) realizują one proces uczenia się sieci neuronowych samodzielnie tzn. rozpoznają relacje występujące w skupieniach poprzez wykrycie wewnętrznej struktury i kategoryzują je w procesie samouczenia. SOM służy do uformowania odwzorowania z przestrzeni wielowymiarowej do przestrzeni jednowymiarowej lub dwuwymiarowej. Główną cechą SOM jest to, że tworzy on nieliniową projekcję wielowymiarową kolektora danych na regularnej, niskowymiarowej (zwykle 2D) sieci. Na wyświetlaczu klastrowanie przestrzeni danych, jak również relacje metryczno-topologiczne elementów danych, są wyraźnie widoczne. Jeśli elementy danych są wektorami, składniki, których są zmiennymi z określone znaczenie, takie jak deskryptory danych statystycznych lub pomiary, które opisują proces, siatka SOM może być wykorzystana, jako podstawa, na której może znajdować się każda zmienna wyświetlane osobno przy użyciu kodowania na poziomie szarości lub pseudo koloru. Ten rodzaj projekcji został uznany za bardzo przydatny do zrozumienia wzajemnych zależności między zmiennymi, a także strukturami zbioru danych.

EN

The article discusses artificial neural networks (ANN). One of the varieties is the Kohonen network, called the Self Organizing Map (SOM), that perform the learning process of neural networks independently, i.e. they recognize relationships occurring in clusters by detecting an internal structure and categorizing them in the process of self-learning. SOM is used to form mapping from a multidimensional space to a one-dimensional or two-dimensional space. The main feature of SOM is that it creates a non-linear multi-dimensional projection of a data collector on a regular, low-dimensional (usually 2D) network. On the display, data space clustering as well as metric-topological relations of data elements are clearly visible. If the data elements are vectors, the components of which are variables with defined meanings, such as statistical data descriptors or measurements that describe the process, the SOM grid can be used as a basis on which each variable can be displayed separately using gray or pseudo-color coding. This type of projection has been found to be very useful for understanding the interrelationships between variables as well as data set structures.

6

Automated and effective content-based mammogram retrieval using wavelet based CS-LBP feature and self-organizing map

Singh V. P., Srivastava R.

Biocybernetics and Biomedical Engineering

|

2018

|

Vol. 38, no. 1

90--105

EN

In this paper, automated, fast and effective content based-mammogram image retrieval system is proposed. The proposed pre-processing steps include automatic labelling-scratches suppression, automatic pectoral muscle removal and image enhancement. Further, for segmentation selective thresholds based seeded region growing algorithm is introduced. Furthermore, we apply 2-level discrete wavelet transform (DWT) on the segmented region and wavelet based centre symmetric-local binary pattern (WCS-LBP) features are extracted. Then, extracted features are fed to self-organizing map (SOM) which generates clusters of images, having similar visual content. SOM produces different clusters with their centres and query image features are matched with all cluster representatives to find closest cluster. Finally, images are retrieved from this closest cluster using Euclidean distance similarity measure. So, at the searching time the query image is searched only in small subset depending upon cluster size and is not compared with all the images in the database, reflects a superior response time with good retrieval performances. Descriptive experimental and empirical discussions confirm the effectiveness of this paper.

7

Zastosowanie sieci samoorganizującej do identyﬁkacji struktur odmienionych węgla kamiennego

Skiba M., Młynarczuk M.

Prace Instytutu Mechaniki Górotworu PAN

|

2016

|

T. 18, nr 3

61--66

PL

Wewnętrzna budowa węgla, możliwa do obserwacji wyłącznie pod mikroskopem, może wykazywać pewne cechy (takie jak: obecność spękań, struktury kataklastyczne czy mylonityczne), które wpływają na zwiększoną pojemność gazową oraz wskazują na pokład szczególnie zagrożony wyrzutami gazów i skał. Problematyka ta była przedmiotem zainteresowania licznych badaczy, którzy dokonali klasyfikacji węgla odmienionego, wyróżniając różne typu strukturalne takiego węgla. W pracy do identyfikacji poszczególnych struktur zastosowano mapę samoorganizującą (SOM). Może ona posłużyć do ujawnienia takich cech w zbiorze danych, które są często niedostrzegalne w wypadku zastosowania sieci neuronowej uczonej z nauczycielem. Badania wykonane zostały na zdjęciach mikroskopowych, a każdą z analizowanych klas opisano za pomocą 7-wymiarowej przestrzeni cech. Zastosowanie sieci samoorganizującej skutkowało klasyfikacją badanych struktur na poziomie 82% skuteczności.

EN

The internal structure of the coal, observable microscopically only, may have certain features (such as the presence of cracks, cataclastic or mylonitic structures) that affect the increased gas capacity and point to the seams particularly endangered by gas and rock outbursts. The issue was the subject of interest for many researchers who have made a classifi cation of structurally altered coal, distinguishing different types of such coal structure. In this paper, individual structures were identified using self-organizing map (SOM). It can be used to reveal such features in the data set, which are often invisible in the case of the use of neural network learning with a teacher. Tests were performed on microscopic photographs, each of the analyzed grades were described using a 7-dimensional feature space. The use of a self-organizing map resulted in the effectiveness of the classification of these structures at the level of 82%.