Wyniki wyszukiwania - BazTech

1

An attempt to landscape segmentation based on remote sensing LULC classification: a case study of Khmelnytskyi Oblast, Ukraine

Będkowski Krzysztof, Komarova Uliana

Teledetekcja Środowiska

|

2026

|

T. 66

1--20

EN

Satellite imagery provides an effective tool for large-scale analysis and supports physico-geographical regionalization. In this study, we investigated the possibility of delineating the landscape of the Khmelnytskyi Oblast (Ukraine) into spatial units based on the results of unsupervised classification of Landsat 8 imagery. Five land cover classes were identified: bare soil, forests, shrublands, agricultural crops, and water bodies. The study area was subsequently divided into 6 × 6 km grid cells, and the proportion of each land cover class within each cell was calculated. In addition, three landscape metrics were determined: IJI, DIVISION, and SHDI. In total, five variables describing land cover proportions and three additional variables describing the spatial structure of their distribution within each grid cell were obtained. The analyses were conducted in five variants using five, eight, and two variables. To reduce interdependencies between variables and to obtain an alternative representation of the data, principal component analysis (PCA) was applied, transforming the variables into a set of uncorrelated components. The transformed variables were then used to perform spatial segmentation using a selected segmentation algorithm. The approximate number of 17 segments was not treated as an optimization target, but rather as a reference level enabling comparison between the independently obtained segmentation results and the physical-geographical regionalization of Khmelnytskyi Oblast proposed by Herenchuk (1980). Parameter selection aimed to achieve a comparable level of spatial detail rather than reproduce the reference division. The obtained spatial divisions showed similarity to the reference regionalization. The results were influenced both by the segmentation algorithm parameters and by the number of variables used in the analysis. Full agreement between our segmentation and the reference regionalization was not possible because the reference regionalization also considered physical-geographical features not reflected in the structure and texture of the satellite imagery. We demonstrated that segmentation in the proposed approach, especially when supplemented with additional data, can be used for the delineation of spatial landscape units.

PL

Obrazy satelitarne stanowią skuteczne narzędzie do analiz wielkoobszarowych oraz wspierają regionalizację fizycznogeograficzną. W tym studium badaliśmy możliwość wykonania podziału krajobrazu Obwodu Chmielnickiego (Ukraina) na jednostki przestrzenne na podstawie wyników nienadzorowanej klasyfikacji obrazów Landsat 8. Zidentyfikowaliśmy pięć klas pokrycia terenu: gleby bez pokrywy roślinnej, lasy, zakrzaczenia, uprawy rolne oraz wody. Następnie analizowany obszar podzielono na kwadraty o wielkości 6 × 6 km i określono udział każdej z klas pokrycia terenu. Ponadto wyznaczyliśmy trzy metryki krajobrazowe: IJI, DIVISION oraz SHDI. Łącznie uzyskaliśmy pięć zmiennych opisujących udziały form pokrycia terenu oraz trzy dodatkowe zmienne opisujące strukturę przestrzenną rozmieszczenia tych klas w każdym z kwadratów. Analizy prowadziliśmy w pięciu wariantach, z wykorzystaniem pięciu, ośmiu oraz dwóch zmiennych. W celu zmniejszenia współzależności między zmiennymi oraz uzyskania alternatywnej reprezentacji danych zastosowano analizę głównych składowych (PCA), przekształcając zmienne do postaci nieskorelowanych komponentów. Przekształcone zmienne wykorzystaliśmy do wykonania podziału przestrzeni za pomocą wybranego algorytmu segmentacji. Przybliżona liczba 17 segmentów nie była celem optymalizacji, lecz punktem odniesienia umożliwiającym porównanie wyników niezależnie uzyskanej segmentacji z regionalizacją fizyczno-geograficzną obwodu chmielnickiego według Herenchuka (1980). Dobór parametrów miał na celu osiągnięcie porównywalnego poziomu szczegółowości przestrzennej, a nie odtworzenie podziału referencyjnego. Uzyskaliśmy podziały przestrzeni wykazujące podobieństwo do regionalizacji referencyjnej. Wpływ na uzyskiwane wyniki miały zarówno parametry algorytmu segmentacji, jak i liczba zastosowanych zmiennych. Uzyskanie pełnej zgodności między naszą segmentacją a regionalizacją referencyjną nie było możliwe, ponieważ regionalizacja ta uwzględniała również cechy fizycznogeograficzne niewidoczne w strukturze i teksturze obrazu satelitarnego. Wykazaliśmy, że segmentacja w zaproponowanym przez nas ujęciu, a także wzbogacona o inne dane, może być wykorzystywana do wyodrębniania przestrzennych jednostek krajobrazowych.

2

Multithreaded evolutionary detection of hyperspheres in high-dimensional point cloud data

Moryń Maciej, Hoser Paweł, Kubica Bartłomiej, Kozera Ryszard

Advances in Science and Technology. Research Journal

|

2026

|

Vol. 20, no. 3

287--298

EN

Object segmentation in multidimensional data spaces is a pivotal component of modern computational analysis. Frequently, accurate segmentation hinges on the detection and localisation of object boundaries. The targeted objects often exhibit spherical symmetry. This paper introduces an algorithm for the automatic detection of n-dimensional hyperspheres embedded in (n+1)-dimensional Euclidean space. The algorithm utilises an evolutionary computation strategy to estimate hypersphere parameters from extensive point clouds. This method demonstrates notable advantages over traditional approaches such as the Hough transform and active surface models. Preliminary results suggest strong potential of the proposed technique as well as its adaptability to broader classes of hypersurfaces, offering a promising extension for future exploration.

3

Comparison of unsupervised machine learning segmentation algorithms in the analysis of unmanned aerial vehicle – based multispectral crop images

Karpiński Paweł, Rzeczkowski Jakub

Advances in Science and Technology. Research Journal

|

2026

|

Vol. 20, no. 2

94--113

EN

In precision agriculture, the analysis of UAV-based multispectral imagery enables spatial differentiation of crop condition, supporting targeted management decisions. This study compares the performance of two unsupervised segmentation algorithms (K-means and Gaussian Mixture Models) in analyzing RGB images of winter wheat, supported by NDVI-based interpretation. Segmentation was performed on RGB orthomosaics acquired at two phenological stages, followed by NDVI analysis to assign physiological meaning to each segment. The average NDVI per cluster was used to reconstruct NDVI maps and objectively assess vegetation condition within segments. In the early growth stage, segmentation primarily reflected spectral variability in the soil background due to low biomass and weak plant–soil contrast. NDVI analysis revealed that seemingly regular clusters corresponded to bare inter-row soil rather than emerging plants – highlighting the limited diagnostic value of RGB segmentation alone at this stage. In the later growth stage, both algorithms accurately delineated field plots and intra-field variability. Using five clusters, the analysis identified zones ranging from dense, healthy vegetation to bare soil. These results demonstrate that combining RGB-based unsupervised segmentation with NDVI analysis is an effective tool for mapping spatial heterogeneity in mature crops, while offering limited standalone value in early growth stages without additional spectral verification.

4

Supervised learning for lower urinary tract identification with limited and imbalanced training data

Wajman Radosław, Banasiak Robert, Sputowski Maciej, Mosorov Volodymyr

Przegląd Elektrotechniczny

|

2025

|

R. 101, nr 6

98--107

EN

In recent years, the advantages of AI in data processing and analysis have become increasingly evident in the medical field. There has been a rapid growth in the application of AI in clinical medicine, including its use in urinary system disease detection. AI offers the capability to process and utilize diagnostic information, presenting new opportunities for precise and individualized treatment while promoting non-invasive diagnostic and therapeutic approaches. This paper introduces an original approach to addressing the supervised learning problem in convolutional neural network (CNN) models for lower urinary tract identification when confronted with a scarce and imbalanced training dataset. The proposed solution involves significantly expanding the diagnostic urinary bladder dataset by increasing the number of samples through various augmentation strategies. However, this approach also intensifies the computational complexity of AI training, rendering it infeasible to load all training datasets into memory simultaneously. To overcome this challenge, a distributed computing approach has been developed, incorporating dynamic loading of training data alongside programming-level memory optimization.

PL

W ostatnich latach zalety sztucznej inteligencji (AI) w przetwarzaniu danych i diagnostyce medycznej stały się coraz bardziej widoczne. Wpływ na to mają postępy w technologii komputerowej i integracja wielu dyscyplin. Warto zauważyć, że nastąpił szybki wzrost zastosowań AI w medycynie klinicznej, w tym jej wykorzystania w technologiach wykrywania chorób układu moczowego. AI oferuje możliwość przetwarzania informacji diagnostycznych, co stwarza nowe możliwości precyzyjnego, spersonalizowanego leczenia, jednocześnie promując nieinwazyjne podejścia diagnostyczne. W niniejszym artykule przedstawiono oryginalne podejście do rozwiązania problemu uczenia nadzorowanego w modelach sieci neuronowych splotowych (CNN) do identyfikacji dolnych dróg moczowych w przypadku ograniczonego i niezrównoważonego zestawu danych treningowych. Proponowane rozwiązanie obejmuje znaczne rozszerzenie zestawu danych diagnostycznych dotyczących pęcherza moczowego poprzez zwiększenie liczby próbek za pomocą różnych strategii. Jednak podejście to zwiększa również złożoność obliczeniową treningu AI, co sprawia, że niewykonalne jest jednoczesne załadowanie wszystkich zestawów danych treningowych do pamięci. Aby przezwyciężyć to wyzwanie, opracowano podejście rozproszonego przetwarzania, obejmujące dynamiczne ładowanie danych wraz z optymalizacją pamięci.

5

Glaucoma detection with the segmentation of optic disc and cup region using improved u-net and classification algorithm

Krishnamoorthy A., Logeswari S., Sivanantham S., Akshaya V., Ganagavalli K., Mohanraj V.

Przegląd Elektrotechniczny

|

2025

|

R. 101, nr 5

71--76

EN

A major contributor to irreversible blindness worldwide, glaucoma affects millions of people each year. To stop vision loss, early detection through precise diagnosis is crucial. In this work, we offer a hybrid deep learning approach for glaucoma prediction that combines ensemble learning approaches with convolutional neural networks (CNNs). The method combines the use of MobileNet V2 and ResNet-18 for classification with an enhanced U-Net model that incorporates residual connections and attention mechanisms for image segmentation. To increase prediction accuracy, our hybrid system makes use of the advantages of both optic cup-disc segmentation and CNN-based classifiers. To maintain computational efficiency and compliance with data privacy requirements, the model is optimized through the use of sophisticated techniques like as AdamW, cyclic learning rate schedulers, and stochastic weight averaging. In the present paper, we have analyzed the various performance metrics like accuracy, sensitivity, specificity, dice coefficient, Jaccard index, F1 score for the conventional methods, and Res-U-Net Architecture (Improved U-Net). Here, Res-U-Net Architecture (Improved U-Net) achieves an accuracy of 0.93 by 80% of the training data. Hybrid deep learning approach for glaucoma prediction that combines ensemble learning approaches with convolutional neural networks (CNNs)

PL

Jaskra, która jest głównym czynnikiem przyczyniającym się do nieodwracalnej ślepoty na całym świecie, dotyka milionów ludzi każdego roku. Aby zapobiec utracie wzroku, kluczowe znaczenie ma wczesne wykrycie poprzez precyzyjną diagnozę. W tej pracy oferujemy hybrydowe podejście do głębokiego uczenia się do przewidywania jaskry, które łączy podejścia do uczenia zespołowego z konwolucyjnymi sieciami neuronowymi (CNN). Metoda łączy w sobie wykorzystanie MobileNet V2 i ResNet-18 do klasyfikacji z ulepszonym modelem U-Net, który obejmuje szczątkowe połączenia i mechanizmy uwagi do segmentacji obrazu. Aby zwiększyć dokładność predykcji, nasz system hybrydowy wykorzystuje zalety zarówno segmentacji tarczy optycznej, jak i klasyfikatorów opartych na CNN. Aby utrzymać wydajność obliczeniową i zgodność z wymogami prywatności danych, model jest optymalizowany dzięki zastosowaniu zaawansowanych technik, takich jak AdamW, cykliczne harmonogramy szybkości uczenia się i stochastyczne uśrednianie wag. Tutaj przeanalizowaliśmy różne wskaźniki wydajności, takie jak dokładność, czułość, swoistość, współczynnik kostki, indeks Jaccarda, wynik F1 dla metod konwencjonalnych i architektura Res-U-Net (Improved U-Net). W tym przypadku architektura Res-U-Net (Improved U-Net) osiąga dokładność 0,93 przy 80% danych treningowych. Hybrydowe podejście do głębokiego uczenia się do przewidywania jaskry, które łączy podejścia do uczenia zespołowego z konwolucyjnymi sieciami neuronowymi (CNN).

6

Application of YOLO and U-Net models for building material identification on segmented images

Voronkov Ruslan, Bezuglyi Mykhailo

Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska

|

2025

|

T. 15, nr 2

13--17

EN

This paper is devoted to the analysis of existing convolutional neuralnetworks and experimental verification of the YOLO and U-Netarchitectures for the identification and classification of building materials based on images of destroyed structures. The aim of the study is to determinethe effectiveness of these models in the tasks of recognising materials suitable for reuse and recycling. This will help reduce construction wasteand introduce a more environmentally friendly approach to resource management. The study examined several modern deep learning models for image processing, including Faster R-CNN, Mask R-CNN, FCN (Fully Convolutional Networks), and SegNet. However, the choice was made on the YOLOand U-Netarchitectures. YOLO is used for fast object identification in images, which allows for quick detection and classification of building materials, and U-Netis used for detailed image segmentation, providing accurate determination of the structure and composition of building materials. Each of these models has been adapted to the specific requirements of building materials analysis in the context of collapsed structures. Experimental results have shown that the use of these models allows achieving high accuracy of segmentation of images of destroyed buildings, which makes them promising for usein automated resource control systems.

PL

Niniejszy artykuł poświęcony jest analizie istniejących konwolucyjnych sieci neuronowych i eksperymentalnej weryfikacji architektur YOLOi U-Net do identyfikacji i klasyfikacji materiałów budowlanych na podstawie obrazów zniszczonych konstrukcji. Celem badania jest określenie skuteczności tych modeli w zadaniach rozpoznawania materiałów nadających się do ponownego wykorzystania i recyklingu. Pomoże to zmniejszyć ilość odpadów budowlanych i wprowadzić bardziej przyjazne dla środowiska podejście do zarządzania zasobami. W badaniu przeanalizowano kilkanowoczesnych modeli głębokiego uczenia do przetwarzania obrazu, w tym Faster R-CNN, Mask R-CNN, FCN (Fully Convolutional Networks) i SegNet, jednak wybór padłna architektury YOLO i U-Net. YOLO służy do szybkiej identyfikacji obiektów na obrazach, co pozwala na szybkie wykrywanie i klasyfikację materiałów budowlanych, a U-Net służy do szczegółowej segmentacji obrazu, zapewniając dokładne określenie struktury i składu materiałów budowlanych. Każdyz tych modeli został dostosowany do specyficznych wymagań analizy materiałów budowlanych w kontekście zawalonych konstrukcji.Wyniki eksperymentów wykazały, żezastosowanie tych modeli pozwala osiągnąć wysoką dokładność segmentacji obrazów zniszczonych budynków, co czynije obiecującymi do wykorzystania w zautomatyzowanych systemach kontroli zasobów.

7

A Method for Segmenting FLIR Images of Maritime Objects Based on Visual Perception

Pietkiewicz Tadeusz

Biuletyn Wojskowej Akademii Technicznej

|

2025

|

Vol. 74, No. 1

17--35

EN

In this paper, a modified adaptive grayscale image segmentation method based on human psychovisual phenomena (VPS) is proposed. This method generates a binarization threshold for each pixel separately. Compared to the0classical VPS method, the use of a pixel observation model in the form of concentric circles is proposed. In the final part of the paper, this method was applied to segment preprocessed FLIR (forward looking infra-red) images of maritime objects. The results obtained by this method were compared with the segmentation results of the same images by the Otsu method using mathematical criteria. The study confirmed better segmentation performance using the modified VPS method.

PL

W artykule zaproponowano zmodyfikowaną adaptacyjną metodę segmentacji obrazów w skali szarości opartą na zjawiskach psychowizualnych człowieka (VPS). Metoda ta generuje próg binaryzacji dla każdego piksela osobno. W porównaniu do klasycznej metody VPS zaproponowano zastosowanie modelu obserwacji pikseli w postaci koncentrycznych okręgów. W końcowej części artykułu metoda ta została zastosowana do segmentacji wstępnie przetworzonych obrazów FLIR (forward looking infra-red) obiektów morskich. Wyniki uzyskane tą metodą zostały porównane z wynikami segmentacji tych samych obrazów metodą Otsu przy użyciu kryteriów matematycznych. Badanie potwierdziło lepszą wydajność segmentacji przy użyciu zmodyfikowanej metody VPS.

8

Analysis on grading of lung nodule images with segmentation using u-net and classification with Convolutional Neural Network Fish Swarm Optimization

Sudha R., Uma Maheswari K.M.

Biocybernetics and Biomedical Engineering

|

2025

|

Vol. 45, iss. 1

90--104

EN

Lung malignant tumors are abnormal growths of cells in the lungs that have the potential to invade nearby tissues and spread to other parts of the body. Early detection of these malignant lung tumors is crucial to avoid complications and improve patient outcomes. However, manual processing consumes time and is a tedious process. This might result in poor estimation on cancer-prognosis, leading the patients into a higher risk of mortality. Many existing literatures have detected the malignant tumors, yet, found certain difficulties with the identification of size, appearance and spread of cancerous-cells in lung region to determine how far it has been occupied. Hence, the present study aims to overcome the existing complications through Deep Learning based Swarm Intelligence Algorithms. Implementation of the proposed work is involved with three stages such as preprocessing, segmentation and classification. Besides, CT scan possess the capability for giving a comprehensive view than X-rays. Data are collected from LIDC-IDRI (Lung Image Database Consortium-Image Database Resource Initiative) with lung CT-images and accomplishes pre-processing by removing noise efficiently using wiener filter. Further, changes in soft tissues of lungs are identified and segmented in the subsequent phase using U-Net and finally classification is performed using CFSO (Convolutional Neural Network Fish Swarm Optimization) to overcome the slight chance of misclassification error as proposed CFSO can lead to more efficient computational processes since FSO algorithms are designed to minimize computational costs while maximizing performance through their metaheuristic nature. This efficiency is particularly beneficial when dealing with large datasets typical in medical imaging, allowing faster processing times without sacrificing accuracy. Hence, amalgamation of CFSO can reduce the number of features, thus speeding up training and inference times. Through the performance assessment, IoU (Intersection over Union) value attained through the analysis is found to be 0.7822. Further, accuracy obtained by the proposed model is 97.80%, recall is 98.49%, precision is 96.8% and F1-score is 97.32%. Findings of the study exhibits the purposefulness of the study in clinical settings by potentially reducing false negatives in lung cancer screening, ultimately improving patient survival rates through earlier detection and treatment.

9

Bladder volume estimation based on USG images

Mosorov Volodymyr, Baradziej Daniel, Chodyka Marta

International Journal of Electronics and Telecommunications

|

2024

|

Vol. 70, No. 4

879-886

EN

The article explores deep learning models in urological diagnostics to measure urinary bladder volume from medical images. It addresses the shortcomings of traditional methods by introducing advanced imaging techniques for more objective and precise analysis. The research employs Convolutional Neural Networks (CNNs) and the MONAI platform for image segmentation and analysis, using data from The Cancer Imaging Archive to focus on urological regions. Findings suggest these models enhance diagnostic accuracy but also highlight the need for further modifications to tailor them to specific medical data, underscoring machine learning's significant role in accurate medical assessments for urology.

10

MultiTumor Analyzer (MTA-20–55): A network for efficient classification of detected brain tumors from MRI images

Sahoo Akshya Kumar, Parida Priyadarsan, Panda Manoj Kumar, Muralibabu K., Mohanty Ashima Sindhu

Biocybernetics and Biomedical Engineering

|

2024

|

Vol. 44, no. 3

617--634

EN

Brain cancer, one of the leading causes of mortality worldwide, is caused by brain tumors. Early diagnosis of tumors and predicting their progression can help doctors to save lives. In this article, we have designed an automated approach for locating and classifying tumors from MRI images. The novelties of the research work include the following two stages: Developing an encoder-decoder type 20-Layered deep neural network (DNN) named MultiTumor Analyzer (MTA-20) with 15 down-sampling layers and 4 up-sampling layers, the segmentation is performed in the initial stage. Here, we have adhered a Leaky ReLU activation function instead of ReLU which learn a parameter with negative values that may have valuable information which is essential specifically for image segmentation. Further, a 55-layered DNN using multistage feature fusion is developed in the second stage of the work for the classification of localized tumors. The classification is performed using developed MultiTumor Analyzer (MTA-55) DNN with Softmax classifier. The efficacy of the designed network is validated using highly cited quantitative measures such as accuracy, sensitivity, specificity, dice similarity coefficient (DSC), precision, and F1-measure. It is observed that the proposed MTA-20 DNN attains the average accuracy, sensitivity, specificity, DSC, and precision of 99.2 %, 94.6 %, 99.3 %, 88 %, and 82.5 % respectively against seven state-of-the-art techniques. Also, it is found that, the proposed MTA-55 DNN provides the overall accuracy, recall, specificity, F1-measure, precision, and DSC of 99.8 %, 99.633 %, 99.844 %, 99.659 %, 99.689 %, and 99.656 % respectively as compared to thirteen state-of-the-art techniques. These results corroborate the superiority of the proposed technique.

11

Segmentation of multigradation images based on spatial connectivity features

Timchenko Leonid, Kokriatskaia Natalia, Tverdomed Volodymyr, Stetsenko Oleksandr, Kaplun Valentyna, Kolesnytskyj Oleh, Reshetnik Oleksandr, Smailova Saule, Zhunissova Ulzhalgas

Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska

|

2023

|

T. 13, nr 3

47--50

EN

The article aims to study the multi-level segmentation process of images of arbitrary configuration and placement based on features of spatial connectivity. Existing image processing algorithms are analyzed, and their advantages and disadvantages are determined. A method of organizing the process of segmentation of multi-gradation halftone images is developed and an algorithm of actions according to the described method is given.

PL

Artykuł ma na celu zbadanie procesu wielopoziomowego segmentacji obrazów o dowolnej konfiguracji i rozmieszczeniu w oparciu o cechy łączności przestrzennej. Przeanalizowano istniejące algorytmy przetwarzania obrazu oraz określono ich zalety i wady. Opracowano metodę organizacji procesu segmentacji wielogradacyjnych obrazów półtonowych i przedstawiono algorytm działań zgodnie z opisaną metodą.

12

Classification and detection of skin disease based on machine learning and image processing evolutionary models

Bordoloi Dibyahash, Singh Vijay, Kaliyaperumal Karthikeyan, Ritonga Mahyudin, Jawarneh Malik, Kassanuk Thanwamas, Quiñonez-Choquecota Jose

Computer Assisted Methods in Engineering and Science

|

2023

|

Vol. 30, no. 2

247--256

EN

Skin disorders, a prevalent cause of illnesses, may be identified by studying their physical structure and history of the condition. Currently, skin diseases are diagnosed using invasive procedures such as clinical examination and histology. The examinations are quite effective and beneficial. This paper describes an evolutionary model for skin disease classification and detection based on machine learning and image processing. This model integrates image preprocessing, image augmentation, segmentation, and machine learning algorithms. The experimental investigation makes use of a dermatology data set. The model employs the machine learning methods: the support vector machine (SVM), the k-nearest neighbors (KNN), and random forest algorithms for image categorization and detection. This suggested methodology is beneficial for the accurate identification of skin disease using image analysis. The SVM algorithm achieved an accuracy of 98.8%. The KNN algorithm achieved a sensitivity of 91%. The specificity of KNN was 99%.

13

PCcS-RAU-Net: Automated parcellated Corpus callosum segmentation from brain MRI images using modified residual attention U-Net

Chandra Anjali, Verma Shrish, Raghuvanshi A.S., Kuber Bodhey Narendra

Biocybernetics and Biomedical Engineering

|

2023

|

Vol. 43, no. 2

403--427

EN

Background: The Corpus callosum (Cc) in the cerebral cortex is a bundle of neural fibers that facilitates inter-hemispheric communication. The Cc area and area of its sub-regions (also known as parcels) have been examined as a biomarker for cortical pathology and differential diagnosis in neurodegenerative diseases such as Autism, Alzheimer’s disease (AD), and more. Manual segmentation and parcellation of Cc are laborious and time-consuming. The present work proposes a novel work of automated parcellated Cc (PCc) segmentation that will serve as a potential biomarker to study and diagnose neurological disorders in brain MRI images. Method: In this perspective, the present work aims to develop an automated PCc segmentation from mid-sagittal T1- weighted (w) 2D brain MRI images using a deep learning-based fully convolutional network, a modified residual attention U-Net, referred to as PCcS-RAU-Net. The model has been modified to use a multi-class segmentation configuration with five target classes (parcels): rostrum, genu, mid-body, isthmus and splenium. Results: The experimental research uses two benchmark MRI datasets, ABIDE and OASIS. The proposed PCcS-RAU-Net outperformed existing methods on the ABIDE dataset with a DSC of 97.10% and MIoU of 94.43%. Furthermore, the model’s performance is validated on the OASIS and Real clinical image (RCI) data and hence verifies the model’s generalization capability. Conclusion: The proposed PCcS-RAU-Net model extracts essential characteristics such as the total area of the Cc (TCcA) to categorize MRI slices into healthy controls (HC) and disease groups. Also, sub-regional areas, Cc1A to Cc5A, help study atrophy progression for early diagnosis.

14

A Comparative Analysis of Image Segmentation Using Classical and Deep Learning Approach

Plaksyvyi Arsen, Skublewska-Paszkowska Maria, Powroźnik Paweł

Advances in Science and Technology. Research Journal

|

2023

|

Vol. 17, no. 6

127--139

EN

Segmentation is one of the image processing techniques, widely used in computer vision, to extract various types of information represented as objects or areas of interest. The development of neural networks has influenced image processing techniques, including creation of new ways of image segmentation. The aim of this study is to compare classical algorithms and deep learning methods in RGB image segmentation tasks. Two hypotheses were put forward: 1) “The quality of segmentation applying deep learning methods is higher than using classical methods for RGB images”, and 2) “The increase of the RGB image resolution has positive impact on the segmentation quality”. Two traditional segmentation algorithms (Thresholding and K-means) were compared with deep learning approach (U-Net, SegNet and FCN 8) to verify RGB segmentation quality. Two resolutions of images were taken into consideration: 160x240 and 320x480 pixels. Segmentation quality for each algorithm was estimated based on four parameters: Accuracy, Precision, Recall and Sorensen-Dice ratio (Dice score). In the study the Carvana dataset, containing 5,088 high-resolution images of cars, was applied. The initial set was divided into training, validation and test subsets as 60%, 20%, 20%, respectively. As a result, the best Accuracy, Dice score and Recall for images with resolution 160x240 were obtained for U-Net, achieving 99.37%, 98.56%, and 98.93%, respectively. For the same resolution the highest Precision 98.19% was obtained for FCN-8 architecture. For higher resolution, 320x480, the best mean Accuracy, Dice score, and Precision were obtained for FCN-8 network, reaching 99.55%, 99.95% and 98.85%, respectively. The highest results for classical methods were obtained for Threshold algorithm reaching 80.41% Accuracy, 58.49% Dice score, 67.32% Recall and 52.62% Precision. The results confirm both hypotheses.

15

Integration of feature extraction, attribute combination and image segmentation for object delineation on seismic images

Khayer Keyvan, Hosseini Fard Esmail, Roshandel Kahoo Amin, Soleimani Monfared Mehrdad, Ahmadyfard Alireza

Acta Geophysica

|

2023

|

Vol. 71, no. 1

275--292

EN

Automatic geological interpretation, specifically modeling salt dome and fault detection, is controversial task on seismic images from complex geological media. In advanced techniques of seismic interpretation and modeling, various strategies are utilized for combination and integration different information layers to obtain an image adequate for automatic extraction of the object from seismic data. Efficiency of the selected feature extraction, data integration and image segmentation methods are the most important parameters that affect accuracy of the final model. Moreover, quality of the seismic data also affects confidence of the selected seismic attributes for integration. The present study proposed a new strategy for efficient delineation and modeling of geological objects on the seismic image. The proposed method consists of extraction specific features by the histogram of oriented gradients (HOG) method, statistical analysis of the HOG features, integration of features through hybrid attribute analysis and image classification or segmentation. The final result is a binary model of the target under investigation. The HOG method here modified accordingly for extraction of the related features for delineation of salt dome and fault zones from seismic data. The extracted HOG parameter then is statically analyzed to define the best state of information integration. The integrated image, which is the hybrid attribute, then is used for image classification, or image segmentation by the image segmentation method. The seismic image labeling procedure performs on the related seismic attributes, evaluated by the extracted HOG feature. Number of HOG feature and the analyzing parameters are also accordingly optimized. The final image classification then is performed on an image which contains all the embedded information on all the related textural conventional and statistical attributes and features. The proposed methods here apply on four seis mic data examples, synthetic model of salt dome and faults and two real data that contain salt dome and fault. Results have shown that the proposed method can more accurately model the targets under investigation, compared to advanced extracted attributes and manual interpretations.

16

DE-XRT coal preparation image overlapping and adhesion particle segmentation method

He Lei, Wang Shuang, Guo Yongcun

Physicochemical Problems of Mineral Processing

|

2022

|

Vol. 58, iss. 6

art. no. 155840

EN

Accurate segmentation of dual-energy X-ray transmission (DE-XRT) coal and gangue image regions are a prerequisite for feature extraction, identification, localization, and separation. A watershed algorithm based on multi-grayscale threshold segmentation (MGTS) is proposed to mark the foreground for the adhesion and overlap of coal and gangue. The grayscale images of foreground objects are segmented using multiple grayscale thresholds, and the number of connected domains is recorded each time. As the gray threshold value decreases, overlapping and adhering objects are gradually separated. The binary image segmented at the grayscale threshold with the most significant number of connected domains is used as a marker region. This marker region is used as the seed point of the watershed algorithm to find the dividing line. The experimental results show that the segmentation accuracy is 91.35%, and the segmentation accuracy of overlapping adhesions of 2, 3, and 4 targets is higher than 90%.

17

On the applicability of integrated UAV photogrammetry and automatic feature extraction for cadastral mapping

Ajayi Oluibukun Gbenga, Oruma Emmanuel

Advances in Geodesy and Geoinformation

|

2022

|

Vol. 71, no. 1

art. no. e19, 2022

EN

The applicability of integratedUnmannedAerialVehicle (UAV)-photogrammetry and automatic feature extraction for cadastral or property mapping was investigated in this research paper. Multi-resolution segmentation (MRS) algorithm was implemented on UAVgenerated orthomosaic for mapping and the findings were compared with the result obtained from conventional ground survey technique using Hi-Target Differential Global Positioning System (DGPS) receivers. The overlapping image pairs acquired with the aid of a DJI Mavic air quadcopter were processed into an orthomosaic using Agisoft metashape software while MRS algorithm was implemented for the automatic extraction of visible land boundaries and building footprints at different Scale Parameter (SPs) in eCognition developer software. The obtained result shows that the performance of MRS improves with an increase in SP, with optimal results obtained when the SP was set at 1000 (with completeness, correctness, and overall accuracy of 92%, 95%, and 88%, respectively) for the extraction of the building footprints. Apart from the conducted cost and time analysis which shows that the integrated approach is 2.5 times faster and 9 times cheaper than the conventional DGPS approach, the automatically extracted boundaries and area of land parcels were also compared with the survey plans produced using the ground survey approach (DGPS) and the result shows that about 99% of the automatically extracted spatial information of the properties fall within the range of acceptable accuracy. The obtained results proved that the integration of UAVphotogrammetry and automatic feature extraction is applicable in cadastral mapping and that it offers significant advantages in terms of project time and cost.

18

Graph-based segmentation with homogeneous hue and texture vertices

Ngo Lua, Han Jae-Ho

Optica Applicata

|

2021

|

Vol. 51, nr 4

541--549

EN

This work presents an automated segmentation method, based on graph theory, which processes superpixels that exhibit spatially similarities in hue and texture pixel groups, rather than individual pixels. The graph shortest path includes a chain of neighboring superpixels which have minimal intensity changes. This method reduces graphics computational complexity because it provides large decreases in the number of vertices as the superpixel size increases. For the starting vertex prediction, the boundary pixel in first column which is included in this starting vertex is predicted by a trained deep neural network formulated as a regression task. By formulating the problem as a regression scheme, the computational burden is decreased in comparison with classifying each pixel in the entire image. This feasibility approach, when applied as a preliminary study in electron microscopy and optical coherence tomography images, demonstrated high measures of accuracy: 0.9670 for the electron microscopy image and 0.9930 for vitreous/nerve-fiber and inner-segment/outer-segment layer segmentations in the optical coherence tomography image.

19

Image segmentation method of mine pass soil and ore based on the fusion of the confidence edge detection algorithm and mean shift algorithm

Jin Feng, Zhan Kai, Chen Shengjie, Huang Shuwei, Zhang Yuansheng

Gospodarka Surowcami Mineralnymi

|

2021

|

T. 37, z. 4

133--152

EN

In the execution of edge detection algorithms and clustering algorithms to segment image containing ore and soil, ore images with very similar textural features cannot be segmented effectively when the two algorithms are used alone. This paper proposes a novel image segmentation method based on the fusion of a confidence edge detection algorithm and a mean shift algorithm, which integrates image color, texture and spatial features. On the basis of the initial segmentation results obtained by the mean shift segmentation algorithm, the edge information of the image is extracted by using the edge detection algorithm based on the confidence degree, and the edge detection results are applied to the initial segmentation region results to optimize and merge the ore or pile belonging to the same region. The experimental results show that this method can successfully overcome the shortcomings of the respective algorithm and has a better segmentation results for the ore, which effectively solves the problem of over segmentation.

PL

W procesie algorytmu wykrywania krawędzi ufności i algorytmu grupowania do segmentacji obrazu zawierającego rudę i glebę, obraz rudy o bardzo podobnych cechach tekstury nie może być skutecznie segmentowany, gdy oba algorytmy są używane osobno. W pracy zaproponowano nowatorską metodę segmentacji obrazu opartą na połączeniu algorytmu wykrywania krawędzi ufności i algorytmu zmiany średniej, który integruje kolor, teksturę i cechy przestrzenne obrazu. Na podstawie wstępnych wyników segmentacji uzyskanych przez algorytm segmentacji zmiany średniej informacja o krawędziach oryginalnego obrazu jest wyodrębniana za pomocą algorytmu wykrywania krawędzi opartego na stopniu ufności, a otrzymane wyniki są stosowane do początkowych wyników segmentacji obszaru w celu optymalizacji i scalenia rudy lub gleby należących do tego samego obszaru. Wyniki eksperymentalne pokazują, że metoda ta może skutecznie przezwyciężyć wady odpowiedniego algorytmu i daje lepsze wyniki segmentacji dla rudy, co dobrze rozwiązuje problem nadmiernej segmentacji.

20

Segmentation of aggregate and asphalt in photographic images of pavements

Mejía Angela Marcela, Alzate Marco Aurelio, Reyes-Ortiz Oscar Javier

Engineering Transactions

|

2021

|

Vol. 69, iss. 1

19--42

EN

Particle size distribution of aggregate in asphalt pavements is used for determining important characteristics like stiffness, durability, fatigue resistance, etc. Unfortunately, measuring this distribution requires a sieving process that cannot be done directly on the already mixed pavement. The use of digital image processing could facilitate this measurement, for which it is important to classify aggregate from asphalt in the image. This classification is difficult even for humans and much more for classical image segmentation algorithms. In this paper, an expert committee approach was used, including classical adaptive Otsu, k-means vector quantization over a set of 8 principal components obtained from 26 features, and a Gaussian mixture model whose parameters are estimated through the expectation-maximization algorithm. A novel cellular automata approach is used to coordinate these expert opinions. Finally, a simple heuristic is used to reduce sub- and over-segmentation. The segmentation results are comparable to those obtained by a human expert, while the sieve size of the segmented images corresponds very well with that obtained from the sieving process, validating the proposed method of segmentation. The results show that with the digital imaging procedure it was possible to detect particles with a size of 100 m with 90% of success with respect to time-consuming manual techniques. In addition, with these results it is possible to establish the homogeneity of the sample and the distribution of the particles within the asphalt mixture.