The paper describes the image superresolution method with aggregate divergence matrix and automatic detection of crossover.Formulation problem, building extreme optimization task its solution for solving automation determination crossover coefficient is presented.Different ways oversampling images algorith ms based on proposed are shows.Based practical experiments shows effectiveness procedure automatically coefficient.Experimentally established procedures at high zoo m resolution by...

In the paper method of single-frame image super-resolution based on singular decomposition matrix operator convergence square is proposed. The characteristic vectors-features are obtained by using Moore-Penrose pseudoinverse operator, which used for enlarged synthesis. series computational experiments images with fluctuation intensity function performed. comparison results others methods have confirmed effectiveness developed approach. main advantages proposed different enlargement...

The new learning-based image super-resolution method is described in this article. process of increasing the resolution video frames or images from a set according to based on weight coefficients synaptic connections. These are obtained by learning neural-like structure pair low and high resolution. dimension influence training generalization properties investigated. comparison work effectiveness proposed existing ones analyzed.

This paper presents a novel algorithm for detection certain types of emergencies relating to fire, smoke and explosions by processing the data recorded from camera monitoring, based on cascaded approach. First, combination Adaboost Local binary pattern (LBP) are using getting Region Interest (ROI) reducing time complexity. Next, alleviate common problems vulnerable such as false positive, we propose use Convolutional Neural Network (CNN). The final experimental results showed that accuracy...

Techniques to detect the flame at an early stage are necessary in order prevent fire and minimize damage.The detection technique based on physical sensor has limited disadvantages detecting early.This paper presents results of using local binary patterns for solving flames problem proposes modifications improve quality detector work.Experimentally found that support vector machines classifier with a kernel Gaussian radial basis functions shows best compared other SVM cores or k-nearest neighbors.

This thesis is devoted to development of visitor's queue density analysis and registration method for a retail videosurveillance systems. Developed foreground segmentation based on initial background modeling, selective temporal median filter local binary patterns. Based the literature review, problem statement has been examined investigated. Basic methods their limitations were also analyzed. Software application, that was developed implement test results shows increases an objects interest...

This article considers two algorithms which are designed for taking into consideration such events as the light changing in video frame, background micro-movements and occlusion of an object. These determine negative influences on segmentation process, because they often arise eyeshot surveillance cameras that operated outdoors.

Neuro-fuzzy models have a proven record of successful application in finance. Forecasting future values is crucial element decision making trading. In this paper, novel ensemble neuro-fuzzy model proposed to overcome limitations and improve the previously successfully applied five-layer multidimensional Gaussian its learning. The solution allows skipping error-prone hyperparameters selection process shows better accuracy results real life financial data.

The paper describes the method image superresolution from two frames on basis of aggregate divergence matrix elements theory and genetic algorithms. Shows different ways for building oversampling images algorithms based proposed method. Experimentally established effectiveness procedures at high zoom resolution by developed Comparative performance evaluation with existing ones.

In parallel with technological development the problem of fraud detection is becoming more and important. Increasing number electronic transactions in various business environments, on one hand, software technology development, other lead to an active increase crime. paper hybrid system machine learning for solving tasks anomalies has been proposed. This consists two subsystems - subsystem (based unsupervised learning) interpretation anomaly type supervised system). The advantage proposed...

In this paper we propose architecture of hybrid generalized additive neuro-fuzzy system. Such system is the Wang-Mendel and models Hastie-Tibshirani. Proposed can be used for solving different tasks computational intelligence data stream mining. The results experimental modelling confirm effectiveness simplicity proposed approach in comparison with conventional systems.

There are proposed method of acceleration large string search in database learning management system

In this paper, the hybrid multidimensional wavelet-neuro-system for pattern recognition tasks is proposed. Also learning algorithm all its parameters (synaptic weights, centers, and widths of wavelet activation functions) based on cross entropy cost function was The proposed system characterized by high speed approximation properties in comparison with well-known approaches. efficiency approach has been justified different benchmarks real data sets.

Time series forecasting can be a complicated problem when the underlying process shows high degree of complex nonlinear behavior. In some domains, such as financial data, processing related time-series jointly have significant benefits. This paper proposes novel multivariate hybrid neuro-fuzzy model for tasks, which is based on and generalizes with consequent layer multi-variable Gaussian units its learning algorithm. The distinguished by separate block each output, tuned respect to output...