Wyniki wyszukiwania - BazTech

1

Local binary pattern defect recognition approach for the friction stir welded AA 1200 and AA 6061-T6 aluminum alloy

Mishra Akshansh

Journal of Mechanical and Energy Engineering

|

2020

|

Vol. 4, No 1

27--32

EN

The research reported in this paper focuses on the application of local binary patterns (LBPs) for surface defects detection. The surface defection detection algorithm for friction stir welded aluminum plates is the key part of the entire surface defect recognition system. Two different grades i.e AA 1200 and AA 6061 plates were similarly joined with the help of Friction Stir Welding process. Python codes for the proposed algorithm were executed on Google Colaboratory platform. The results obtained prove that the local binary patterns method can be used for real-time surface defects detection in friction stir welded joints.

2

Detection of Modic changes in MR images of spine using local binary patterns

Athertya Jiyo S., Saravana Kumar Gurunathan, Govindaraj Jayaraj

Biocybernetics and Biomedical Engineering

|

2019

|

Vol. 39, no. 1

17--29

EN

Background and objective: With increase in prevalence of lower back pain, fast and reliable computer aided methods for clinical diagnosis associated with the same is needed for improving the healthcare reach. The magnetic resonance images exhibit a change in signal intensity on the vertebral body close to end plates, which are termed as Modic changes (MC), and are known to be clear indicators of lower back pain. The current work deals with computer aided methods for automating the classification of signal changes between normal and degenerate cases so as to aid physicians in precise and suitable diagnosis for the ailment. Methods: In order to detect Modic changes in vertebrae, initially the vertebrae are segmented from sagittal MR T1 and T2 imaged using a semi automatic cellular automata based segmentation. This is followed by textural feature extraction using Local Binary Patterns (LBP) and its variants. Various classifiers based on machine learning approaches using Random Forest, kNN, Bayes and SVM were evaluated for its classification performance. Since medical image dataset in general have bias towards healthy and diseased state, data augmentation techniques were also employed. Results: The implemented method is tested and validated over a dataset containing 100 patients. The proposed framework achieves an accuracy of 81% and 91.7% with and without augmentation of data respectively. A comparative study with the state of art methods reported in literature shows that the method proposed in better in terms of computational cost without any compromise on classification accuracy. Conclusion: A novel approach to identify MC in vertebrae by exploiting textural features is proposed. This shall assist radiologists in detecting abnormalities and in treatment planning.

3

Formulation and statistical evaluation of an automated algorithm for locating small bowel tumours in wireless capsule endoscopy

Jagadeesan A., Sivaraman J.

Biocybernetics and Biomedical Engineering

|

2018

|

Vol. 38, no. 4

782--793

EN

Wireless capsule endoscopy (WCE) is an imaging modality which is highly reliable in the diagnosis of small bowel tumors. But locating the frames carrying tumors manually from the lengthy WCE is cumbersome and time consuming. A simple algorithm for the automated detection of tumorous frames from WCE is proposed in this work. In the proposed algorithm, local binary pattern (LBP) of the contrast enhanced green channel is used as the textural descriptor of the WCE frames. The features employed to differentiate tumorous and nontumorous frames are skewness (S) and kurtosis (K) of the LBP histogram. The threshold value of the features which offers the trade-off between sensitivity and specificity is identified through Receiver Operating Characteristic (ROC) curve analysis. At the optimum threshold, both the features exhibited a sensitivity of 100% and specificity of 90%. The skewness and kurtosis of the LBP computed from the enhanced green channel of tumorous and nontumorous frames differ significantly ( p « 0.05) with a p-value of 2.2 x 10-16. The proposed method is helpful to reduce the time spent by the doctors for reviewing WCE.

4

Emotion recognition from facial images using binary face relevance maps

Herud T., Kawulok M., Smolka B.

Studia Informatica

|

2015

|

Vol. 36, nr 4

29--41

EN

This paper is focused on automatic emotion recognition from static grayscale images. Here, we propose a new approach to this problem, which combines a few other methods. The facial region is divided into small subregions, which are selected for processing based on a face relevance map. From these regions, local directional pattern histograms are extracted and concatenated into a single feature histogram, which is classified into one of seven defined emotional states using support vector machines. In our case, we distinguish: anger, disgust, fear, happiness, neutrality, sadness and surprise. In our experimental study we demonstrate that the expression recognition accuracy for Japanese Female Facial Expression database is one of the best compared with the results reported in the literature.

PL

W artykule tym przedstawiono zagadnienie rozpoznawania emocji na podstawie obrazów w skali szarości. Prezentujemy w nim nowe podejście, stanowiące połączenie kilku istniejących metod. Obszar twarzy jest dzielony na mniejsze regiony, które są wybierane do dalszego przetwarzania, z uwzględnieniem binarnych map istotności. Z każdego regionu ekstrahowany jest histogram lokalnych wzorców binarnych, a następnie histogramy są składane do wektora cech i klasyfikowane za pomocą maszyny wektorów podpierających. W naszym przypadku rozróżniamy takie emocje, jak: gniew, wstręt, strach, szczęście, neutralność, smutek i zaskoczenie. Podczas naszych eksperymentów pokazaliśmy, że nasze podejście umożliwia poprawę skuteczności rozpoznawania emocji dla bazy Japanese Female Facial Expression względem innych istniejących metod.

5

A texture-based method for classification of schizophrenia using fMRI data

Pouyan A. A., Shahamat H.

Biocybernetics and Biomedical Engineering

|

2015

|

Vol. 35, no. 1

45--53

EN

This paper presents a texture-based method for classification of individuals into schizophrenia patient and healthy control groups based on their resting state functional magnetic resonance imaging (R-fMRI) data. In this research a combination of three different classifiers is proposed for classification of subjects into predefined groups. For all fMRI scans, the number of time points is reduced using principal component analysis (PCA) method, which projects data onto a new space. Then, independent component analysis (ICA) algorithm is used for estimation of the independent components (ICs). ICs are sorted based on their variance. For feature extraction a texture based operator called volume local binary patterns (VLBP) is applied on the estimated ICs. In order to obtain a set of features with large discrimination power, a two-sample t-test method is used. Finally, a test subject is classified into patient or control group using a combination of three different classifiers based on a majority vote method. The performance of the proposed method is evaluated using a leave-one-out cross validation method. Experimental results reveal that the proposed method has a very high accuracy.

6

Optimal classification method for smiling vs neutral facial display recognition

Nurzyńska K., Smołka B.

Journal of Medical Informatics & Technologies

|

2014

|

Vol. 23

87--94

EN

Human face depicts what happens in the soul, therefore correct recognition of emotion on the basis of facial display is of high importance. This work concentrates on the problem of optimal classification technique selection for solving the issue of smiling versus neutral face recognition. There are compared most frequently applied classification techniques: k-nearest neighbourhood, support vector machines, and template matching. Their performance is evaluated on facial images from several image datasets, but with similar image description methods based on local binary patterns. According to the experiments results the linear support vector machine gives the most satisfactory outcomes for all conditions.